1 I B ^ARY 
 
 OF THE, 
 U N IVL R.S'ITY 
 OF ILLINOIS 
 
The person charging this material is re- 
 sponsible for its return on or before the 
 Latest Date stamped below. 
 
 Theft, mutilation and underlining of books 
 are reasons for disciplinary action and may 
 result in dismissal from the University. 
 
 UNIVERSITY OF ILLINOIS LIBRARY AT URBANA-CHAMPAIGN 
 
 
 * / ► ,: :* 
 
 *. *- 
 
 BtMi 
 
 i 
 
 w« 
 
 < 
 
 3 !'C 
 JSE OftLY 
 
 1982 
 
 L161— O-1096 
 
Digitized by the Internet Archive 
 in 2013 
 
 http://archive.org/details/designconceptsfo203mcco 
 
X 
 
 MATHEMATICS 
 
 Report No. 203 
 
 DESIGN CONCEPTS FOR AN INFORMATION RESOURCE CENTER 
 
 with option of 
 An Attached Automated Laboratory 
 
 by 
 Bruce H. McCormick 
 and 
 A. M. Richardson, Architectural Design 
 
 May 1, 1966 
 
Return this book on or before the 
 Latest Date stamped below. 
 
 Theft, mutilation, and underlining of books 
 are reasons for disciplinary action and may 
 result in dismissal from the University. 
 University of Illinois Library 
 
DESIGN CONCEPTS FOR AN INFORMATION RESOURCE CENTER 
 
 with option of 
 An Attached Automated Laboratory 
 
 by 
 Bruce H. McCormick 
 and 
 A. M. Richardson, Architectural Design 
 
 March 15, 1966 
 
 Department of Computer Science 
 University of Illinois 
 Urbana, Illinois 61803 
 
TABLE OF CONTENTS 
 
 1. Introduction 
 
 2. Applications 
 
 3- List of Associated Facilities 
 
 h. Building 6 
 
 5. Microform File/ Display (with R. C Amendola) • • 22 
 
 6 . Computer Hardware 37 
 
 7 • Programming System ^5 
 
 8- Communications Net 55 
 
 9- Automated Laboratory 62 
 
 10. Summary 67 
 
 Appendix A: Subcontracting and Cost Considerations. 72 
 
 Acknowledgments 78 
 
 Page 
 
 1 
 
 3 
 
 h 
 
1. INTRODUCTION 
 
 An automated Information Resource Center, properly conceived, 
 creates for the user a new environment: a setting with information sources 
 to saturate his ability to learn, to formulate new questions, and to 
 record his responses. Here the user can immerse himself in pertinent 
 communication with data acquisition facilities; pattern recognition and 
 interpretation algorithms; techniques for the data processing of natural, 
 visual and machine-oriented languages; audio/visual display; and 
 supplementary reference material. The creation of this responsive environment 
 must be the primary service performed by the resource center. 
 
 To design the information resource center, we can in part extrapolate 
 from present experience in working l) with audio-visual linkage systems, 
 2) with disk storage reservoirs for instructional and research materials, 
 and 3) with man-machine dialogue systems. We must provide centralized 
 distribution for the decentralized use of book and non-book materials. We 
 must develop procedures which allow the user (institutional, professional or 
 business) to monitor data acquisition at remote laboratories, to experiment 
 freely with a rapidly evolving pool of programmed algorithms, and to be kept 
 intimately aware of related research reported elsewhere. 
 
 It would be a mistake, however, to emphasize exclusively the 
 technological side of an information resource center - the collecting, 
 classifying and dissemination of information. Rather, our broader challenge 
 must remain to stimulate a new human responsiveness, a widening horizon of 
 our " minds eye ", as we move from the quasi-static world of movable type to 
 the evanescent worlds of the computer-generated display. Emerging here 
 is a new humanism -- thought and action by automata centered upon distinctly 
 human interests. For how can we model systems of the complexity of the 
 human nervous system; regulate the weather; develop a nationwide data base 
 for the science of medicine, except by the creation of information processing 
 systems of this complexity? How can we prepare an indigenous population 
 to abate the rigors of a computer- dominated society except by the creation 
 of a network of centers of this type, plainly visible, freely available 
 to the public at moderate cost, and directed to their use, enlightenment 
 and continuing education? 
 
 -1- 
 
The goal espoused here is adventurous. As programmed respondent, 
 we will be asked to chart a parallel and supplementary course to the 
 processing of information by the user, with only such nuances as he 
 manages to convey as a guide. Our success in this venture will depend 
 critically upon the dedication and talent of those individuals who chose 
 to program and edit the new epistemology to emerge as common usage of 
 information resource centers augments more traditional strategies for thought 
 
 -2- 
 
2 . APPLICATIONS 
 
 One modular Information Resource Center design, with only minor 
 architectural modifications, could find these diverse applications in 
 the institutional, professional and business worlds: 
 
 1) Experimental Automated Library for a university, junior college 
 or technical institute 
 
 2) Data Acquisition and Interpretation Center for a research 
 laboratory 
 
 3) Medical Information Center for a hospital, clinic or medical 
 
 arts group -- with option of attached clinical pathology laboratory 
 
 k) Com put er Center for a junior college or small university -- or 
 one of several for a larger university 
 
 5) Control Exchange for a network of remote consoles 
 
 6) Center for Computer Aided Instruction 
 
 T) Profe ssional Information Exchange for architects, doctors, 
 attorneys, etc. 
 
 ^ ) M ilitary Command and Control Center 
 
 9) Des ign Automat ion C enter for engineering/ drafting/ graphic arts 
 
 10 ) Marketing, P rocess Control and Distributi on Center for a 
 manufacturer/ distributor 
 
 H) Res ervat ions and Tr affic Con trol Center for transportation 
 industry 
 
 -3- 
 
3- LIST OF ASSOCIATED FACILITIES 
 
 The Information Resource Center building should be designed as a 
 core module compatible with its use in a larger architectural complex. The 
 types of supplementary space required by the eleven proposed applications 
 of the Information Resource Center are summarized below: 
 
 1) Automated Laboratory Space 
 
 high energy physics laboratory 
 life sciences laboratory 
 electronics laboratory 
 pilot production plant 
 
 2) Professional Office Space 
 
 system programming offices 
 
 medical arts center 
 
 aerial photo interpretation center 
 
 biomedical image processing laboratory 
 
 merchandizing mart/ commodities exchange 
 
 law offices 
 
 architectural/engineering design offices 
 
 3) Instructional Space/ Auditorium 
 
 to include: 
 
 learning laboratories 
 
 convertible classroom/ seminar areas 
 
 auditorium 
 
 audio-visual control rooms 
 
 map room 
 
 k) Library Space 
 
 browsing or undergraduate library 
 link to large research library 
 lounge/ reading rooms 
 
 5) Small Special Purpose Laboratory 
 
 clinical pathology laboratory 
 
 -k- 
 
integrated circuits laboratory- 
 special purpose photographic scanning/processing laboratory 
 
 6) Communications Center 
 
 to include: 
 
 television, radio and film studios 
 
 audio-visual workshop 
 
 graphic arts workshop 
 
 control rooms 
 
 rehearsal-observation areas 
 
 offices 
 
 7) Motel-type Space 
 
 restaurant/motel for information processing center 
 extended medical care building 
 
 8) Manufacturing Plant 
 
 typical low cost one story manufacturing shell 
 
 9) Custom Laboratory Installation 
 
 reactor, accelerator, missile site, etc. 
 
 10) Microwave Tower/Cable Raceways 
 
 for information links to other centers, both local and 
 remote 
 
 11) Landscaping/Parking 
 
 -5- 
 
k . BUILDING 
 
 i+.l Space Requirements 
 
 We can picture a typical Information Resource Center as a two- 
 story structure containing 50 individual carrels with display console, 
 microform disk file for 100,000 books, computer area, and operations/ 
 administration space. Outside an optional tower for microwave link to 
 special remote users and similar centers at a distance can be provided. 
 
 i+.l.l Display Console Carrels 
 
 Carrels, approximately 6' x 6' in floor area, are to be provided 
 the individual user. Each carrel is equipped with a console display (CRT 
 or other) on-line to the computer complex of the Information Resource 
 Center. A typical installation might contain 50 carrels. 
 
 Display consoles are restricted to four types: 
 
 1.1 Console A (for reading and study): 
 
 molded glass viewing screen and adjacent keyboard 
 
 forming a self contained panel, which is attached 
 directly to the architectural wall. 
 
 1.2 Console B (for design automation/ graphic arts): 
 
 a display surface like a drafting table with facilities 
 (light pen, television camera, etc.) for input of 
 graphic arts information. Nearby keyboard matches 
 that of a type A console. 
 
 1-3 Console C (for data acquisition and interpretation): 
 mostly used remotely for data acquisition, with 
 cart-mounted TV camera, monitor and LINC-type computer 
 interface attributes. 
 
 l.k Console D (for seminar and small group discussion): 
 very similar to type A console, except designed to 
 accommodate a party of 2-k persons in intimate 
 discussion. 
 
 All consoles are provided with individual lighting, ventilation, 
 acoustical insulation, work surfaces, chair(s) and privacy curtain. To 
 
 W. A. Clark and C E. Molnar, The LINC: A Description of the Laboratory 
 Instrument Computer, Annals of the New York Academy of Sciences, Vol. 115, 
 Art. 2, July 31, 196*4. _£_ 
 
enhance acoustical isolation of the carrels, it is recommended that the 
 floor in this area be carpeted. 
 
 U.1.2 Microform Disk File 
 
 Isolated from the carrel area but readily seen by the visitor 
 are glass enclosed columns almost three stories high containing the 
 microform disk file units of the Information Resource Center. In a 
 typical installation, storage for 100,000 books would be provided on these 
 optical disks. 
 
 To be specific we might visualize this microform disk file as 
 provided by 6k record (i.e., microform disk) changing mechanisms, each 
 5' x 2' x 1 1/2'. (A modification of commercially available juke-box 
 mechanisms would do admirably. ) Each unit of this capacity might store 
 1600 books. This material is stored photographically. That is, a visual 
 image of any page (or page segment) can be read out onto a high resolution 
 closed circuit display/ storage tube at a console. Using this changer 
 mechanism as a building block, a double walled structure could be built in 
 two layers of such blocks, two units long and sixteen units high to make 
 a total of 6k such boxes in the system. 
 
 In general, though 6k optical disk file modules might be considered 
 typical, systems ranging from 10 to 1000 modules can be contemplated. 
 Accordingly the outer housing for the disk file should be considered a part 
 of the architectural design. A possible disk housing design - for 
 architectural integration - suggests precast concrete skeleton with provisions 
 for file module inserts. In any case the ease of inserting or removing 
 a microform disk file module for maintenance purposes must be considered. 
 The double walled column, or the hollow cylinder, arrangement has the added 
 maintenance feature that extraction and insertion of file modules can be 
 done in isolation from the public. The design illustrated overpage features 
 an inexpensive service elevator in the space enclosed by the two walls of 
 file modules. 
 
 ^-.1.3 Computer Room 
 
 The computer room contains the central processors, fast storage, 
 backup storage and local input/output of the computer system. Also the 
 
 -7- 
 
central exchange service of the telecommunications net is placed in this 
 
 area • 
 
 For purposes of a preliminary estimate we assign sizes to each 
 
 of these elements: 
 
 Width x Depth x Height 
 
 3.1 Central processor and fast storage 16 ' x 3' x 5' 
 
 3.2 Backup storage 16 ' x 3' x 5' 
 
 3.3 Local Input/ Output 2U' x 3* x 5' 
 3.U Telephone Central Exchange 6' x 3' x 5' 
 
 We will assume all units can be maintained from the front. On this basis 
 a computer room area of oO' x 20' is minimal but possible. 
 
 Provision should be provided for a maintenance technician; 
 suggesting an adjoining area for oscilloscope and combination workbench/ 
 desk. 
 
 An alternative design places the input/output equipment immediately 
 adjoining the routing room area discussed below. In any case the flow of 
 input/ output media between the local input/ output equipment and the routing 
 room area is critical. 
 
 An additional cost consideration concerns the floor in the computer 
 area. The needs here are two: i) space for cable raceways, and ii) an 
 air plenum for equipment intake air. One solution, conventional but 
 expensive, is to lay a false floor over the entire area. An alternative 
 proposal is to lay a prefabricated pattern of combination plenum trenches 
 and cable raceways directly into the foundation concrete. In either case 
 the air conditioning burden of 20-30 tons maximum is carried by the building 
 system and not by the installed equipment. 
 
 Finally there is the added complication that users and visitors 
 often have a keen interest to see what the computer looks like. The computer 
 should be on display through glass. It is not recommended however that 
 visitors be permitted to enter the area. 
 
 U.1.1+ Administration/ Operations Area 
 
 Space set aside for administration and operations could vary widely 
 from one installation to another. 
 
 -8- 
 
Minimal requirements for administration include a 2-man business 
 office. In addition, office space for four system consultants should be 
 provided. This latter resident advisory staff would serve as consultants 
 on the programming system. Many of these inquiries might be addressed to 
 a member of this staff using console-to-console communication. Accordingly 
 at least two display consoles, possibly one of type A and one of type B 
 should be available for consultation with potential new users. It is 
 convenient then if this office space is directly accessible from the entrance 
 lobby and near the routing room area. 
 
 A routing room area, immediately in evidence upon entering the 
 lobby of the building, serves as the "mailroom" of the center. Here are 
 housed facilities for introducing new documents into the system: by film, 
 conventional or video tape, closed circuit television, etc. Also immediately 
 accessible here are the devices for generating output media: xerographic 
 printout, film, etc. It is convenient if a subdivision of this area serves 
 as an information office where introductory printed literature, course 
 descriptions, manuals, reports, etc. can be obtained. An area approximately 
 20' x 12' would be desirable for the routing/ information room. 
 
 An area for microfilming and photographic reproduction is a 
 considerable convenience. Lower cost basement space can be used for this 
 purpose. Photo reduction of large layout sheets should be possible, using 
 equipment such as the Keuffel and Esser Micro-Master camera/projector 
 system for engineering drawings. Total space requirements might be kept 
 to 16' x 30' . 
 
 Storage space for supplies, largely paper, must be provided. A 
 floor area of 10' x 20' should prove adequate. 
 
 Finally the application may require additional space, say 
 16' x 28', for equipment special to the installation. A center associated 
 with a high energy physics laboratory may wish to provide this space for 
 precision film scanners. Or a stereo coordinatograph and large scale digital 
 plotter for topographical maps might fit the needs of an aerial photo 
 interpretation laboratory. When the space requirements exceed this one room 
 it is assumed that the application warrants independent building space. This 
 area, properly speaking, can be viewed as the seed area from which new 
 
 -9- 
 
applications of the Information Resource Center are grown. 
 
 1+.1.5 Overhead Space 
 
 Conventionally assigned to this space are corridors, restrooms, 
 mechanical equipment, etc (Corridor space within the carrel area, 
 however, will be charged to that area - see preliminary design below). 
 
 As mentioned above, the entrance to the center opens into a 
 lobby, with immediate access to the routing/ information office, to the 
 resident advisory staff, and probably to the business office. The lobby 
 should also serve as a gallery for arousing interest in the work of the 
 center. 
 
 k.2 Desirable Features 
 
 Ideally these Information Resource Centers should serve as a 
 cultural and intellectual stimulant to society. Architectural design is 
 important if the Information Resource Center is to be provided at minimum 
 cost a building that is both functional and aesthetically elevating. 
 In part, what is required here is the industrial design of a man-machine 
 complex, for economics in the fabrication of a center can only be achieved 
 if the building costs, as also the computer, microform disk file, and 
 programming development costs, are spread over a large number of similar 
 centers. 
 
 4.2.1 Quick Construction 
 
 To minimize construction costs the use of local building trades must 
 be curtailed. This calls for modular prefabricated construction erected either 
 upon a basement roughed in by the local building trades or directly upon a 
 slab. 
 
 A suggested goal is that construction and equipment installation 
 should take no more than 60 days upon preparation of the unfinished basement or 
 prepared slab. In particular work to be performed by the local trades, 
 prior to the delivery of the prefabricated components, should make no reference 
 to computers and require only the most conventional type construction. 
 
 ■10- 
 
4.2.2 Package Units 
 
 At one extreme the industrial design of a Information Resource 
 Center could define one uniform "product", as is commonly done in 
 designing gasoline stations. This direction appears too restrictive, 
 and does not allow flexibility to meet the varied applications of these 
 centers . 
 
 A related approach is to develop a set of package units which 
 in a tinker-toy way allows economical construction of a wide class of 
 buildings- This point of view suggests the "carrel wall", discussed below, 
 as one building module. 
 
 4.2.3 Cable Raceways 
 
 Attention must be given to adequate cable raceways between the 
 display consoles of the system and the computer. In like manner coaxial 
 cable raceways should be provided for trunks of the communications net to 
 remote users. 
 
 Rapid installation calls for prefabrication of all wire harnesses, 
 and standardization of these where possible. Ideally wiring would come 
 completely installed in the prefabricated modules. 
 
 4-3 The Carrel Environment 
 
 The user operates in three relevant spheres of influence: 
 
 1. The outside world via remote consoles 
 
 2. The information resource center 
 3- The machine via the console. 
 
 All three spheres must be taken into account, but for this discussion 
 we will consider the user placed in an environment comprised of the 
 building proper, his carrel area and his console. Sphere two is primarily 
 a static situation; console operation in sphere three is a dynamic process 
 involving communication with the machine and will be treated apart (see 
 Section 5-2) and with great care. 
 
 In the carrel environment, the user will want the following: 
 
 -11- 
 
Privacy control 
 
 Sound control 
 
 Airflow and. temperature control 
 
 Lighting control 
 
 Storage space allocation 
 
 A wastepaper disposal 
 
 Control over his seating arrangement 
 
 I+.3.I Privacy 
 
 This is of paramount importance in the information resource center 
 situation. The user must be able to shut himself away from the building 
 area and passageways. He will at times have to work with confidential 
 material (medical records, etc.) and security then becomes important. He 
 will often have personal material, and personal privacy becomes important. 
 In any case, some means to shut out the world must exist for the user. 
 
 U.3.2 Sound 
 
 It is not enough to make the carrels as quiet as possible and 
 hope for the best. Extreme concentration is called for sometimes, and 
 ambient conversation twenty feet away could be very annoying to a person under 
 stress . 
 
 We are therefore charged with allowing him to soundproof himself 
 at will by some means. If the carrel privacy seal is just a thin door or 
 curtain door, a great deal of ambient noise can get through -- unless the 
 door, as carrel, can be adequately soundproofed. 
 
 ^■•3-3 Airflow and Temperature 
 
 It is evident that each carrel will have its own ventilation. The 
 airflow rate should be controllable by the user, partly for his own comfort 
 and partly because it looks, like the only way to also allow him to control 
 the temperature. If he is allowed to change the exchange rate, he can 
 probably alter the carrel temperature by + 10 degrees; this seems adequate. 
 Also smoking fumes can be exhausted through this personal ventilation 
 system -- even with the carrel door open. 
 
 ■12- 
 
^•3'^ Lighting 
 
 This will be a key factor in determining the satisfaction of 
 the user. The lighting level of a room can profoundly affect the 
 emotional attitude and state of awareness of the occupants. There is a 
 great variation in preferred levels of illumination by various people. 
 One absolutely must work in shadowless light at 63OO K. with no lights 
 on instruments; the next wants about four tallow candles sitting on his 
 desk and brightly lighted instrument panels. I submit that is useless 
 to try to set the lighting to a known optimum. Instead one should provide 
 adjustable indirect even illumination for the carrel area, a direct adjustable 
 selective light for the workspace, and illumination control for the console 
 parts themselves. At least these three conditions must be met. 
 
 U.3-5 Storage and Power 
 
 There should be room for a book or two on the workspace, some 
 room for notepapers, a shelf for temporary storage of things weighing a 
 few pounds. There should be one or more 110 VAC power outlets for tape 
 recorders, etc. The workspace should probably have a surface area of six 
 square feet or so. The work surface should be rock steady and should not boom 
 and bang when struck. All surfaces should be smooth, dense, easy to clean 
 and hard to damage. The writing surface should be a light neutral grey. It 
 should not be colored as it is difficult to read tracings or other thin 
 materials over a dark or colored surface. Nor should it be white, as this 
 gets dirty very easily and generally stays that way. There should be a 
 minimal number of level changes, cracks, ledges, etc. 
 
 ^••3-6 Wastepaper 
 
 There should be a wastebasket. 
 
 U.3.7 Seating 
 
 This too is important to a user. He may spend several hours in 
 the confines of the carrel area, and he must be comfortable. The chair should 
 be fully adjustable in all positions. It should have firm cushioning and be 
 very strongly built of high quality materials. It should roll, have arms and 
 a full back. In the case of automated drafting or automated lab work, 
 industrial stools might be a better choice. 
 
 -13- 
 
h.k The Carrel Wall Concept 
 
 In line with the above discussion, A- M. Richardson of Richardson, 
 Severns, Scheeler and Associates/ Architects has suggested as a building module 
 the "carrel wall"- This U-shaped member, with base 6' x 6' and 20' high, 
 serves both as the wall of two carrels -- one on the first floor and one on the 
 second floor directly above -- and as a structural member of the building. 
 These wall segments (external or internal) would be thin precast concrete 
 with thermal insulation core. All ventilation, console brackets, lighting 
 pods, cable raceways, and floor supports are shaped by the concrete casting 
 form. 
 
 This scheme gives great freedom of architectural expression. 
 Exterior carrel walls can be joined by narrow slit windows extending two 
 
 floors, or by conventional brick facing. Indicative of this concept are 
 
 + 
 the model photographs shown below.— 
 
 Copyright and Patent rights are pending. 
 
 -Mr. Richardson's work was supported entirely from private non-governmental 
 funds . 
 
 -Ik- 
 
CD 
 
 h 
 O 
 O 
 
 -P 
 
 w 
 cd 
 
 CD 
 
 Xi 
 
 -p 
 o 
 
 g 
 
 o 
 
 CD 
 -P 
 
 a 
 
 CD 
 
 o 
 
 CD 
 O 
 h 
 
 O 
 w 
 
 0) 
 
 « 
 o 
 
 •H 
 -P 
 
 CO 
 
 p 
 
 o 
 
 H 
 
 ch 
 O 
 
 d 
 
 •H 
 
 CD 
 
 -d 
 a 
 
 CD 
 
 cd 
 u 
 
 •H 
 P-4 
 
 ■15- 
 
-16- 
 
Ill 
 
 
 < 
 a. 
 
 h 
 10 
 
 z 
 ui 
 
 
 
 j 
 < 
 
 
 
 
 J 
 
 | 
 
 Q 
 
 Q 
 
 Q 
 
 
 K 
 
 
 
 i 
 
 J 
 
 
 z 
 y 
 2 
 
 p [ 
 
 J 
 
 
 
 
 
 
 I 
 
 
 
 
 z , 
 
 
 
 
 
 
 
 
 
 
 z - 
 
 
 
 
 
 o h , 
 
 
 
 
 
 - u 
 
 
 1 
 
 
 
 h id 
 
 
 
 
 
 
 -, 
 
 
 
 
 
 
 3 
 
 
 (t 
 
 
 
 a c 
 
 
 X 
 
 
 
 £* 
 
 
 at 
 
 
 
 "■ ui 
 
 
 4 
 
 
 
 0* 
 
 
 Q 
 
 
 
 h 
 
 
 l — "-' 
 
 
 I 
 
 
 1 
 
 
 LiJ 
 
 < 
 d 
 
 
 h 
 1/) 
 
 ■IT- 
 
%;I 
 
 o 
 
 c 
 
 CO 
 U 
 -P 
 
 W 
 
 s 
 o 
 
 p^ 
 
 c 
 
 QJ 
 CD 
 CO 
 
 W 
 
 cd 
 
 cd 
 H 
 •H 
 
 s 
 
 o 
 
 O 
 o 
 
 •H 
 
 a 
 
 (3 
 
 •H 
 
 •H 
 
 w 
 ctf 
 -C 
 9< 
 
 o 
 
 ■H 
 
 fH 
 
 CU 
 
 -p 
 
 H 
 
 Ph 
 
 O 
 O 
 
 -P 
 w 
 
 s 
 
 •H 
 
 ■18- 
 
> 
 
 0) 
 •H 
 
 > 
 
 -P 
 
 o 
 u 
 
 -p 
 
 W 
 
 CO 
 
 H 
 
 <U 
 -P 
 
 o 
 
 cu 
 
 o 
 La 
 
 O 
 
 w 
 
 « 
 o 
 
 •H 
 -P 
 
 O 
 
 LfN 
 
 SO 
 •H 
 
 ■19- 
 
y 
 
 -p 
 
 w 
 CO 
 0) 
 X! 
 -P 
 Jh 
 O 
 
 o 
 
 CD 
 •H 
 
 > 
 
 <u 
 -p 
 c 
 
 CD 
 
 o 
 
 0) 
 
 o 
 
 3 
 
 o 
 
 03 
 CD 
 
 K 
 
 £ 
 O 
 •H 
 
 -P 
 CO 
 
 S 
 R 
 o 
 
 CD 
 
 3 
 
 •H 
 
 -20- 
 
-p 
 
 O 
 
 6 
 o 
 
 0) 
 •H 
 > 
 
 0) 
 
 •p 
 o 
 
 0) 
 
 o 
 
 O 
 W 
 HI 
 K 
 
 a 
 
 o 
 
 •H 
 
 -P 
 ctf 
 
 6 
 R 
 
 o 
 
 S 
 
 3 
 
 DO 
 •H 
 (in 
 
 -21- 
 
5- MICROFORM FILE/ DISPLAY 
 
 5.1 Microform File 
 
 An adequate working library for a major discipline (medicine, 
 engineering, law, life sciences, etc.) requires approximately 100,000 
 books. To this collection must be added special archives, e.g. medical 
 records, appropriate in any one application. 
 
 We are faced then with the problem of accessing and displaying 
 on demand any of several tens, to even hundreds, of million pages in 
 nimimum time. The images displayed can be expected to include print, 
 handwriting, line drawings and photographs. To handle the latter, eight 
 distinguishable shades of gray should be preserved in transmission. These 
 images must be displayed for a length of time ranging from one second to 
 many minutes -- perhaps an hour for difficult material. 
 
 At this stage of technology the only reasonable way to store this 
 material is in microform -- as direct microphotographic images of an 
 optical disk (or plate) file. A warehouse for 100,000 books with conventional 
 human access is a moderate sized library; the building costs here (without 
 books) would be comparable to the anticipated entire cost of an Information 
 Resource Center including computer. If microfilm at conventional reduction 
 ratio (about 2U:l) is used, then the burden of building cost is removed, but 
 
 the cost of books in this form approaches the anticipated cost of the entire 
 
 7 
 center. Furthermore, to provide random access within seconds to 10 or 
 
 greater microfilm images is a formidable engineering task. Accordingly 
 
 to keep material and space costs for storage down, we require microform 
 
 storage using reduction ratios in excess of 100-diameters . 
 
 The general principle of the proposed microform disk (plate) storage 
 was developed about 1962 by the National Cash Register Company. In the 
 variant suggested here, pages to be recorded are first reduced to a high 
 quality 16 mm frame size, one page per frame. Next a matrix of these pages 
 is further reduced, in the step-and-repeat manner, onto a photochromic 
 plate. After inspection, the photochromic glass intermediate is copied 
 by contact printing onto a glass disk (plate) master backed with a high 
 
 -22- 
 
resolution diazo emulsion. A glass permanent master is employed for 
 dimensional stability and resistance to wear. 
 
 Actually an 8 1/2 x 11 inch page format would be recorded in 
 the system proposed here as two micro images -- one for the top and bottom 
 halves respectively. Each 8 1/2 x 5 l/2 inch half page format would then 
 be recorded as a 0.031 inch x 0.020 inch image , a 275:1 linear reduction. 
 Postulating 2 = l6,38U images per disk (plate), we suggest two alternative 
 formats: 
 
 disk: 32 bands of 512 images each on a 6 7/8 inch disk 
 (commercial 7 inch record size). An additional 2 
 innermost bands are provided for indexing and clocking 
 information. For critical dimensions, see Figure 8. 
 plate: 128 rows of 128 images (columns) each on a 3 x 5 inch 
 plate. Additional lines for indexing and fiducial 
 marks for alignment can be provided. 
 
 The reasons for the photochromic intermediate are apparent when 
 we observe that it requires no chemical processing, can be inspected for 
 image quality immediately, can have changes of individual images made at 
 any time, has no grain, is easily capable of resolving 1000 lines per 
 millimeter, has both a low gamma and high contrast, and is re-usable 
 indefinitely. Thus, one may be sure that all l6,38^+ images for a disk 
 (plate) are good before making the permanent master. 
 
 The question now arises of how disks (plates) generated from a 
 complete set of masters, may be stored and accessed in a typical installation. 
 For disk storage the natural retrieval mechanism is one or another of the 
 record changing mechanisms of a commercial juke box. Despite the low cost 
 of these units ($500 each), techniques are available, discussed below, 
 which can be adapted to this type unit. 
 
 For plate storage the natural retrieval mechanism is the carousel 
 arrangement wherein the plates, perhaps 3*5 inches overall, are arrayed 
 much like the Kodak carousel projector. The carousel is indexed to the 
 proper plate -- either stepped or by continuous rotation. An arm extracts 
 
 Less of course for most book page formats 
 
 -23- 
 
/oof/LO' 
 
 x 
 
 
 J 
 
 i 
 
 ^ 
 
 j 
 
 fa 
 
 •', 
 
 K 
 
 Q: 
 
 
 < 
 
 
 
 1 
 
 X 
 w 
 
 •H 
 
 Q 
 
 <u 
 H 
 
 •H 
 
 O 
 <Vh 
 O 
 
 fH 
 CJ 
 •H 
 
 S 
 
 00 
 0) 
 
 •H 
 
 -2U- 
 
the plate, drawing it to its proper column position in the optical path 
 in one motion, and then moving to the proper row. The image is then in the 
 proper position for the micro-optics pickup and is projected upon the 
 vidicon scanning head by flash illumination. (Flash illumination would 
 also be used in the disk format. ) The optical system never moves. After 
 flashing, the plate may remain in the extraction arm if no other user makes 
 a request for another file in the carousel, or it may be re-inserted to 
 search for the next file. 
 
 A typical book can be recorded in a rectangular block of 32 x 32 
 micro images, approximately 1.20 x 0-75 inches. This suggests we 
 reinterpret the microform disk as a circular array of lo of these blocks, 
 and in like manner the 3 x 5 plate as a h x h array of these 1024-image blocks 
 This interpretation suggests an alternate mechanism to access individual 
 images of the disk (plate), using a technique from integrated circuit 
 technology. Here the microcircuit dimensions are typically comparable to 
 those for a micro image, and the silicon wafer corresponds well to the 
 dimensions of a microform block. The suggested image access mechanism 
 uses a "fly's eye" lens in vacuum contact with the microform block. Then 
 any of the 102^ images of the microform block can be projected back to the 
 scanner vidicon by illuminating its field. For this latter purpose a raster 
 of gaseous discharge cells can be used. (See discussion of plasma memory, 
 Section 6). When using this technique the absolute displacement of the 
 micro image from the vidicon tube face is not critical. 
 
 To keep access time reasonable, the extensive microform disk (or 
 plate) file must be composed of many identical file modules, each with one 
 or more image pick-up scanners. A reasonable compromise, suggested by 
 the thorough field testing of the commercial juke box, is to have Gk file 
 modules, each containing 100 optical disks (plates). 
 
 Access time in a typical installation depends both on the 
 retrieval mechanism and the queue of requests for use of the file module. 
 From the time the request for an image is accepted by the retrieval 
 mechanism, the processing times might run as follows: 
 
 + At a disk angular velocity of 33 l/3 rpm a flash of 1 microsecond duration 
 will resolve approximately 3000 TV lines in the micro image. 
 
 -25- 
 
1. Indexing to the proper disk (plate) 
 
 2. Extraction and band (column) select 
 
 3. Angular (row) select 
 
 Total 
 
 2- 5 sec. 
 
 1-2 sec. 
 
 1-2 sec 
 
 ^ - 9 seconds 
 
 Thus the display would appear in k to 9 seconds after the request. Alternate 
 pages of one book (or programmed text) appear with average access time of 
 approximately 1 second. 
 
 5.2 Consoles 
 
 The consoles of the Information Resource Center must feature at 
 least the following items: 
 
 The cathode ray storage and display tube 
 
 Input keyboard 
 
 Light pen input 
 
 Closed circuit television camera input (optional) 
 
 Audio input-output 
 
 Lighting control (readily accessible to the user for 
 
 independent adjustment of room, screen and indicator 
 
 panel levels of illumination) 
 
 A working surface that will accommodate writing materials 
 
 and be large enough to handle an ancillary portable 
 
 typewriter or a tape recorder 
 
 Some form of temporary storage space for user materials 
 9> At least one convenience outlet 
 10. The console must be easy to keep clean, have a minimal 
 
 number of sharp edges, and be pleasant to attend. 
 
 It must be functional but not sterile. 
 
 The user must not be able to satisfy his curiosity about 
 
 the interior of the electronic equipment, and yet this 
 
 equipment must be maintainable from the front. 
 
 A good adjustable chair with arms should be provided. It 
 
 should roll. 
 Ik. The floor should be carpeted and the walls soundproofed. 
 
 7 
 
 8 
 
 11 
 
 12 
 
 13 
 
 These times are based on purely mechanical accessing. If a matrix array of 
 lights is used, these times can be trimmed by about 2 seconds, and the access 
 mechanism is simpler. 
 
 -26- 
 
5-2.1 The Console Environment 
 
 The console itself becomes a whole sphere of operation, an 
 environment for sensory input/ output . Basically there are the 
 following sense and control mechanisms which must be considered: 
 
 1. Visual 
 
 2 . Auditory 
 
 3- Left and right hands 
 k. Feet and knees 
 
 The human is bilaterally symmetrical about a rough vertical. 
 Ours will be mostly seated. Going out from the waist about two feet, the 
 included useful area without twisting is about four square feet. This 
 area is well within the visual range. Assuming that one can work comfortably 
 in such an area at just over elbow height, one has about sixteen inches to 
 shoulder height from the elbow. There is about ^-600 cubic inches of optimum 
 control location space -- a triangular prism 2k inches square l6 inches 
 high. We should attempt to keep all controls within this volume. The 
 most frequently used controls and those used for long periods of time should 
 be in the position one would assume when normally placing his arms on a 
 desk, that is, bent slightly across the body straight at elbow height. 
 
 5-2.1.1 Visual Input/ Output 
 
 The question has risen as to how the visual data should input/ 
 output at the console. Several ways of accomplishing this are possible. 
 Only further testing will settle the question. There is going to be at 
 least one CRT visual output device in the console. If there is only one, 
 it will be large (21") and will operate in several modes. The idea of 
 displaying an upper or lower half -page is appealing. The other half of 
 the tube would then be active for editing purposes with the light pen, or 
 act as a scratch pad. If this mode is used, the upper half of the tube must 
 be the visually active half containing the stored image, and the lower half 
 is the scratch pad. A horizontal pair of images side by side is possible, 
 but not as good as a vertical set. 
 
 If two distinct tubes are used, one should be almost horizontal- 
 faced. The other should be above it at a face angle to vertical of thirty 
 degrees -- and this should be the visually active tube. The scratch pad 
 
 -2 7 - 
 
tube should always approximate a piece of paper placed on a desk. 
 
 A third approach is to use a down-looking closed circuit television 
 
 camera to sense actual written material, a literal sketchpad. This should be 
 placed so that a sheet of paper is located below and slightly to the right of 
 the visual monitor tube. There should be a fixed paper position so that 
 the television pickup can organize the area dependably. It is possible that 
 the arrangement of having two readout CRTs and a television down-looking 
 pickup will be the most flexible and advantageous. 
 
 A recent study of reading requirements for reproduced documents 
 has been reported by Walkup, Ullrich, Stock, and Dugan of the Battelle 
 Memorial Institute.- Working on micro image readers with projected copy, 
 they found roughly the characteristics which follow: 
 
 1. The total image should be book -page size (9 1/2 x 5 1/2 inches) 
 
 2. The letters should be dark on a light background. 
 
 3- The letter image should be about 10-12 point size, stroke 
 medium bold. (One printer's point = l/72 inch). 
 
 k. The majority of letter images should be lower case and 
 not italicized. 
 
 They report the following psychophysical results. 
 
 1. The reader must be able to assume several comfortable 
 positions. 
 
 2. The image optimally should have at least 7 lines per 
 millimeter resolution (optical, not television). 
 
 3- The type characteristic must be like printed matter 
 
 rather than look photo-altered in boldness. 
 h. The reader must be binocular. 
 5- Image contrast should be 0-7 or better, based on 
 
 C = (Bg -Bf)/Bg where C is contrast, Bg is ground 
 
 brightness, Bf = figure brightness. 
 6. Brightness of the ground should be about that of ambient 
 
 roomlight . 
 
 To comment further, although this work was on film readers, since 
 transparent sources were used, it would seem that some of the results apply 
 
 "Notice that this camera can be provided pan, tilt, and zoom controls so that 
 
 + commercial 525-line unit could be used. 
 
 ~~L. E. Walkup, 0. A. Ullrich, J. R. Stock, J. M. Dugan, "The Design of Improved 
 
 Microimage Readers for Promoting the Utilization of Microimages", Proceedings 
 
 of the Eleventh Annual Meeting and. Convention , The National Microfilm Association; 
 
 Annapolis, Maryland, 1962, pp. 283-310- ~~~ 
 
 -28- 
 
to CRT work. Certain contentions are obviously proper. High contrast 
 good resolution, the fact that a black letter reads better on a light 
 ground than vice versa. All of these things hold true for CRT storage 
 tube work. A luminous letter under high contrast conditions tends to 
 "wash out" detail and edge sharpness from an effect known as "illuminance". 
 The page size recommendation fits our notion of half a page rather well. 
 
 There remains only a comment or two about phosphors. We will 
 not worry too much about the color aspects of the CRT display as we are 
 limited as to choice of phosphors on technical grounds. The phosphor 
 should be as near the glass face as possible to limit parallax. There 
 should be a glare elimination device available. Every precaution should 
 be taken to eliminate sources of reflections and glare. Finally the 
 displays should allow the user to vary brightness and contrast to suit his 
 needs. 
 
 5-2.1.2 Audio Input/ Output 
 
 Audio communication presents certain special problems. Audio 
 storage should be remote for all permanent materials. Cartridges of other 
 audio materials can be left at the routing room and enter the system 
 through channels provided there. Accordingly no tape recorder need be 
 provided at the console. 
 
 It is not recommended that conventional headphones be provided as 
 these are unsanitary and troublesome to maintain. This presents a problem 
 in that audio material should be presentable to the user but not blast out 
 into other areas except in the special case of group use. A first obvious 
 solution is to use stethoscope-type headphones and to follow procedures now 
 often used on jet flights. Here the transducer is permanently enclosed in 
 the console and the user couples his sterilizable headphones through a jack 
 in the console. Preferably it may be possible to design a highly directional 
 pair of non-contact speakers that beam the sound directly at the user's 
 head and insure that the sound stays with the user. 
 
 One technique in use in language labs is to have a recorded 
 language instructor speak phrases. The user listens to the voice, presses 
 a button and repeats the phrase which is recorded by a separate remote 
 recorder. He then listens to the instructor followed by his own phrase 
 
 -29- 
 
and compares the two. As such techniques as this undoubtedly find their 
 place in the information resource center, each console must provide the user 
 with immediate access to a remote recorder. 
 
 5.2.1.3 Use of Hands 
 
 The human hands are versatile in the extreme; they should be 
 utilized to their fullest extent at the console. They will play with the 
 controls, the microphones, I/O switches, gains, etc., whatever and whenever 
 control changes are needed. It may be possible to eliminate many of the 
 functional moves the hands must make. We cannot eliminate all of them. 
 Either a light pen or a stylus will be used --at least. Ambient lighting, 
 etc., the second environment sphere, will require control by hands. Such 
 functions as are not eliminated should be distributed in the confines of the 
 optimal 1+600 cubic inch space mentioned above. These should be grouped into 
 relative hierarchies and located according to function sequence, frequency, 
 and period in terms of the human operator. If possible, the controls and 
 switches should be distributed with a slightly higher load for the left 
 hand. This is because the right hand will often be busy "writing". Functions 
 should, if possible, be apparent from the controls, and the controls should be 
 located close to that which is controlled. The reach should not exceed 30" 
 forward in the sitting position, nor above the shoulder, nor below the 
 waist, nor as close as 2" from the waist, nor more than 30 horizontally on 
 either side of the plane through the shoulder normal to the plane of the 
 shoulders. When the functions to be performed are established, it will be 
 possible to give a more definitive set of parameters. 
 
 5-2.1.14 Use of Feet 
 
 Foot switches may be practical in some applications in the 
 information resource center. For that matter, knee switches should also be 
 possible. These are unknowns at present. At any rate, sufficient space 
 for feet and legs to get out of the way should be provided. The equipment 
 must be so designed that the user cannot inadvertently insert his extremities 
 into dangerous areas. Neither should the console have a penchant for 
 snagging nylons • 
 
 A mock-up of the entire console environment will be essential before 
 
 -30- 
 
the final design is set. This mock-up, after initial scale models, will 
 be necessary for checking all electronic functions and human factors. It 
 should be as close in physical characteristics to the real environment as 
 possible. The mock-up should be built in two stages. 
 
 1. Rough mock-up 
 
 2. Finished mock-up 
 
 The rough mock-up should be full size and be non-functioning, but 
 provided with a closed circuit television monitor and remote camera. This 
 setup would be used for dimension checking, evaluating human factors, and 
 be readily alterable. The final mock-up should be fully functional and 
 finished as would be the end product. It should include the carrel, chair, 
 carpeting, functioning CRT storage display and audio system, lighting, sound 
 control, and ventilation. 
 
 In summary, only the best human, engineering criteria should be 
 followed in the console design. We are building not a console, but an 
 environment. A user may spend several hours in this environment and it must 
 be comfortable, functional and undistracting. Quiet is very important; as is 
 a controllable light level both for the CRT and the rest of the console area. 
 The design of this area could well affect the acceptance or rejection of 
 the information resource center by the public. 
 
 5-2.2 The Cathode-Ray Charge-Storage Display Tube 
 
 Having settled on the CRT type of display, we now see that either 
 a long persistence phosphor must be used in the disk material presentation, 
 or some type of buffer memory must be provided to avoid accessing a stored 
 image after the disk has moved. The choice of a long persistence phosphor 
 would be wrong because the display of dynamic images then becomes degenerate. 
 Dual mode phosphors can be had and may be a solution, but scan converters 
 certainly are not the answer. Nor does a digital buffering device seem 
 plausible for the simple reason that we have analog optical pictures to 
 present, and the storage capacity to retain a picture with digitized 
 gray scale would be ridiculous. 
 
 Repeated access to the same image on a disk is a waste of time 
 that may be used more profitably, in particular to service remote users of 
 the center. So it appears that whether the disk or the carousel file 
 
 -31- 
 
arrangement is used, we must transfer photo information as an analog 
 signal to the CRT. Static information should be written once, and 
 independently we should be able to display dynamic material in the usual 
 television raster manner, or economics permitting, directly generate 
 alphanumeric characters. 
 
 The device which comes to mind which would handle such an 
 assignment was developed by Hughes Aircraft and is called the "Multimode 
 Tonotron" . It is basically a CRT storage tube but has some interesting 
 characteristics. It features halftone stored display as in a storage tube 
 but with the added characteristics of selective erasure, high contrast, 
 dynamic presentation if desired, a high write and erase rate, and even 
 color of a sort. It is made in 21" diameter tube formats as of this year. 
 
 Stored resolution, determined at limiting writing speed by the 
 shrinking raster method is listed b,y the manufacturer as 35 lines per 
 inch when writing gun grid no. 2 is operated at +250 volts, and 50 lines per 
 inch when +1000 volts is applied. This resolution, currently available, 
 already reaches the required system performance. 
 
 •32- 
 
5«3 Materials Preparation 
 
 If 100 books per working day were prepared for insertion into 
 the microform file, it still would take four years to build up the 
 100,000 book library. It therefore seems imperative that at least half 
 of the images for the prototype library be ready by the time the first 
 installation is ready to run. As a corollary, file and console display 
 equipment must be settled upon, built, tested and corrected as one of 
 the earlier developmental tasks. 
 
 In what format should a book be transcribed for microform 
 storage: one page -* one image, or should the entire text be re-edited to 
 more suitably fit the requirements of console display? Strangely this 
 central question seems never to have been seriously considered, possibly 
 because the cost of transcription of text to any other format was considered 
 prohibitive . 
 
 At first the suggested re-editing of all text appears radical- 
 On reflection the proposal is seen to have much additional merit. Perhaps 
 first the reader should examine several books or other records such as he 
 would want available in the microform file. The simple comparison of a 
 dictionary page and typewritten copy illustrates the range of image 
 resolution required. Each page of a dictionary, however, can be rather 
 trivially reproduced as several micro images. The fine print of the 
 dictionary represents a publishing compromise: readability versus printing 
 cost and portability of the book. For microform storage the "printing 
 costs" as well as physical size of text are essentially negligible. 
 Accordingly in this new medium the publisher could edit material anew 
 to better suit the man-machine interface at a display console. In particular 
 the use of small fonts at the console must be discouraged -- first because 
 of the file/display system will fail to resolve this degree of detail, and 
 secondly because, unlike a book, the user tends to view a console from a 
 fixed displacement. 
 
 Least we be misunderstood, the re-editing of text for microform 
 display would be organized to use the original text as if it were cut up 
 and repasted together. Most commonly a page of a textbook (say) would be 
 reproduced as two micrc images: one image for the top half page, one image 
 for the bottom half. For intransigent materials, for example text with 
 
 -33- 
 
footnotes in ultra-small fonts, more radical procedures are called for. 
 Specifically let us assume that the text has been microfilmed at one or 
 more reduction ratios such that a U000 line scan will carefully preserve 
 image detail of the type font. The text is then scanned line by line, 
 momentarily encoded a raster of bits (or bit pacs, if gray scale required). 
 These lines are then broken into shorter lines with corrections of 
 justification, hypenation, etc. This new bit pattern is then transmitted 
 to a recording camera. In effect this is a somewhat more sophisticated 
 computer analog of scissors drafting or photocomposition. We assume 
 that each page, as examined, is displayed on a monitor showing both the 
 original and the revised images. And of course a dialogue control language 
 would have to be worked out, but here the experience with computer-generated 
 type setting demonstrates the feasibility of the proposal. With equipment 
 now being completed for the ILLIAC III computer system, we estimate average 
 transcription times of under 10 seconds/page for routine conversion of 
 textbook material, or approximately 2 hours/book for intransigent materials. 
 And of course for common text, where two micro images per page suffice, the 
 format conversion could be done at time of initial microfilming. 
 
 The programming costs to provide the elaborate "apparatus of 
 bibliographic control" - can be justified only if a goal of a hundred centers 
 in the same discipline, e.g. for medical information, is contemplated. 
 Accordingly with this hypothetical hundred centers in mind, let us evaluate 
 two alternative proposals. 
 
 The first proposal requires that every text for submission to the 
 microform store be re-edited (if necessary by a computer-controlled high 
 resolution scanner/ camera system) to allow clear copy at a display console 
 when the micro image is examined under actual working conditions. More 
 specifically we will assume that the video consoles can be provided with a 
 1029 line scan and 30 MC video bandwidth. (in the presently available 
 Granger Associates camera/monitor this corresponds to a vertical resolution 
 of 720 lines and a horizontal resolution of 85O lines.) The video camera 
 
 Alternatively we can directly use the digital type setting tapes, if these 
 exist. 
 
 " Clapp, Verner, Future of the Research Library , University of Illinois Press, 
 Urbana, Illinois, 1964. 
 
 -3^- 
 
examining the micro image is assumed to scan at 60 fields per second with 
 2:1 positive interlace to produce 30 frames/ sec for direct monitoring. 
 
 If the cost of re-editing can be kept to $100/book (which we 
 believe is reasonable and readily attainable) for the 15 per cent of 
 intransigent materials and to $10/book for routine (2 micro images/page) 
 materials, then the additional re-editing cost for 100,000 books is 
 
 ^ 100 dollars >, ,10 dollars v Q _ ___. , . N 
 ^ b^ok 15,000 books) + ( ^—^ x 85,000 books) 
 
 = $2,350,000, or a proportional charge of $23,500 per center, or slightly less 
 than one quarter per book/ center. 
 
 The second alternative hypothesizes a radical improvement in 
 resolution for both the micro image scanner and the console storage/ display 
 system. With this idealized system adequate resolution is hypothesized 
 to reproduce an adequate likeness of essentially any book page. Books are 
 transcribed into microform by direct one page to one image conversion. Even 
 if such equipment does exist, say three years hence, it can realistically be 
 assumed that the cost of this equipment will be more, harder to maintain, 
 etc Even the optical system of the file system must be matched to a higher 
 performance level requiring tighter mechanical tolerances. The incremental 
 cost per display console or disk (plate) file module would be certainly in 
 excess of $1000 each. Therefore assuming 100 installations with 50 consoles 
 and 50 active file modules, we compute 
 
 t ™ . , , . . , m console/file modules 
 
 100 installations x 100 : ttt^ — T~- 
 
 installation 
 
 x $10 oo incremental cost = $ 10 ,000,000 
 equipment 
 
 We conclude that for 100 installations there is no economic justification 
 for not pre-editing the text and relieving the resolution burden required of 
 the file/ display equipment. Above 100 installations the arguments for 
 pre-editing are enhanced. 
 
 Consider momentarily the source of documents in the microform 
 files. For instance what might be the reaction of a publisher to the use 
 of his material in a microform file? There are many aspects to this 
 
 ■35- 
 
question: copyright laws, proper remuneration, varieties of commercially 
 competing microform files, etc. 
 
 Our view of the situation is as follows: until perhaps 1975 
 use of materials in microform will be sufficiently limited, and normally 
 backed up by ancillary books, to pose no serious threat to the conventional 
 publishing industry. The optimal form of a microform file will probably 
 also be an open question during this same period. During this period file 
 standards will most likely be set by the monopolistic position of one 
 information storage and retrieval system manufacturer --as related computer 
 standards were set during this past decade. However by the end of the decade 
 a thriving microform publishing industry will have emerged, following patterns 
 not unlike those followed by the audio record industry. 
 
 To play safe then a publishing house could provide any, or better, 
 all the competing microform file companies with appropriately edited text 
 on a l6 mm microfilm master, or alternatively on microfiche cards. There 
 might be no charge for this material ; it would be treated as promotional 
 material. In return the information resource center using the microform 
 material remunerates the publisher by providing a sales function: a 
 user of the center can request a personal copy of any text in the file and 
 charge this to his credit at the center. The computer compiles daily the 
 orders for each publisher and places the order. In effect the center then 
 serves a dual function as a book store, where the customer can conveniently 
 order personal copies of books by browsing through the files of the center. 
 
 The situation for university libraries and other institutional 
 libraries prohibited from commercial sales is more difficult. First there 
 is the problem of the book sold, and probably written, predominantly for the 
 institutional library trade. Here there might be a direct charge for the 
 microform material -- quite as for the same material in book form. And 
 as for other materials the technicality of merely listing but of not actually 
 ordering the texts for the customer may suffice legally. 
 
 In any case we have considered here the entry of newly published 
 materials, as technical obsolescence makes the use of older materials 
 of minor importance within a span of five years. 
 
 -36. 
 
6. COMPUTER HARDWARE 
 
 6 .1 General 
 
 A computer that can "see" patterns is required by the Information 
 Resource Center. The ability to scan millions of microform file images, 
 material presented over closed circuit television, photos, maps, graphs, 
 engineering drawings, experimental film and microscope slides--and interpret 
 this material faster than heretofore possible--def ines the type of computer 
 required. The machine must be capable of reading pictures fed to it and 
 of doing the analysis necessary to detect patterns of interest to the user. 
 For example, it will be required to digitally encode microform text and 
 figures so that indexing and fact retrieval processes can be automated. 
 In conjunction with an attached clinical pathology laboratory, the machine 
 might be programmed to detect certain diseased cells on a microscope slide, 
 and to prepare a printed and annotated copy of the result. 
 
 Existing commercial computers being used to process pictures perform 
 this task rather inefficiently since they were not designed for it. Picture 
 processing is becoming increasingly important in a number of areas, making 
 it desirable to design computers to do the job efficiently. Potential 
 applications of this type machine range from library automation, preparation 
 of instructional materials, design automation and graphic arts automation 
 to the processing of scientific data: bubble chamber and spark chamber 
 photographs of nuclear physics, to the electron and optical microscope 
 images of the life sciences. 
 
 ILLIAC III is a pioneering development in this area. This 
 experimental computer using a new concept of "parallel" organization is 
 being designed and constructed at the University of Illinois under Atomic 
 Energy Commission contract. 
 
 ILLIAC III, also known as the Illinois Pattern Recognition Computer, 
 is being built to test the parallel design concept for automatic recognition 
 of similar patterns in a series of pictures or images which normally would 
 require the close scrutiny of human eyes. The computer will be able to 
 perform up to 102^- logical operations at once. It is expected to be the 
 first of a new generation of computers which depart from today's "one step 
 
 -37- 
 
at a time" concept of computer design. 
 
 The experimental computer is an outgrowth of studies begun in 
 1961 on systems to scan bubble chamber photographs of high energy particle 
 tracks after collision "events" and convert them into more digestible 
 numerical data. By late 1963 the scope of the contract was expanded to 
 emphasize pattern recognition without reference to a specific scientific 
 discipline. 
 
 The advanced design of ILLIAC III has also attracted the interest 
 of the Advanced Research Projects Agency (ARPA) of the Department of Defense. 
 ARPA is adding a 65,536 word core memory, a disk back up store, and a closed 
 circuit television communication system to this prototype computer system. 
 These additions will permit users with only occasional data-processing 
 demands to have access to the central pattern recognition computer from 
 several remote consoles located around the Illinois Campus, or beyond. 
 
 Fabrication of the machine is well under way at the Department of 
 Computer Science at the University of Illinois, of which Professor John R. 
 Pasta is the head. The project director for ILLIAC III is Professor 
 Bruce H. McCormick. 
 
 On the basis of this experience, the computer proposed for the 
 Information Resource Center is a second-generation ILLIAC III, retaining the 
 order code and system design of this machine but taking full advantage of 
 recent technological advances. 
 
 6.2 Constituents 
 
 Subsystems of the proposed computer consist of 
 
 1.1 Central processors 
 
 1.2 Fast storage 
 1-3 Backup storage 
 1-k Local input/ output 
 
 Each of these components is discussed in more detail below. 
 
 6.3 Central Processors 
 
 The basic computer architecture will follow the organization of 
 the Illinois Pattern Recognition Computer (ILLIAC III) in broad outline. Basic 
 
 -38- 
 
processors will include arithmetic units, pattern articulation units, exchange, 
 taxicrinic processors, input/output processors, channel interface units 
 and appropriate links to the (remote) communications net. All these processors 
 (units) have an ILLIAC III prototype. 
 
 A basic change contemplated is to redesign for fully integrated circuit 
 implementation. This redesign, however, would transcend current practice -- 
 with an eye to increased operating speed and lower manufacturing cost. The 
 logical design, largely following ILLIAC III logical drawings, would be 
 adapted to techniques where after testing "wallpaper" patterns of 500-1000 
 logical elements are interconnected (by sputtering and evaporation) directly 
 on the silicon wafer — i.e. before dicing the wafer into functional electronic 
 blocks. Several semiconductor companies (Motorola, Texas Instruments) have L, 
 nascent development efforts in this area and have indicated interest in applying 
 these techniques to a full scale computer system. 
 
 A related change is the processor redundancy to be employed. 
 Considerable scientific and industrial investment might depend upon reliable 
 2k hour, 365 days/year operation. For data acquisition uses (see Se tion 9: 
 Automated Laboratory) it appears reasonable to assume the machine must be 
 maintained on line. That is, no scheduled down time can be permitted. 
 Accordingly at least two of every processor unit must be an integral part of 
 the minimum machine complex. Every processor (unit) must be provided with a 
 programmable disconnect feature -- initiated upon repeated subsystem malfunction. 
 Maintenance testing, both preventative and diagnostic post mortem, of all 
 processors must be considered a part of the normal computational burden. 
 
 The exchange system of ILLIAC III, however, should be redesigned 
 for greater redundancy and error isolation. This task would be necessary 
 in any case if the number of processing units is to be increased, as suggested 
 by redundancy arguments. 
 
 Of greater programming significance is the change contemplated for 
 the pattern articulation unit design. The planes of this unit would still be 
 partitioned into fast and slow registers. Integrated circuit technology 
 would be used for the fast logic ("stalactites" in ILLIAC III parlance) and M0S 
 technology for the planes of the backup storage ("transfer memory"). A plane 
 size of 6k x 6k is contemplated, unless a breakthrough in technology would 
 permit direct implementation of 128 x 128 or even 256 x 256 size arrays. 
 
 -39- 
 
6.U Fast Storage 
 
 Storage, as in the ILLIAC III, will be organized by 10 bit chare 
 (9-bit information byte + parity bit). The setup time of gates within the 
 exchange net suggests that the common interface between processors and fast 
 memory modules through the exchange net still use the 100-bit single rail 
 system (5 characters INBUS, 5 characters OUTBUS). 
 
 Of fast storage designs, magnetic thin film technology is now 
 being given the greatest industrial development. Large memories with 250 
 nanosecond cycle time have been designed for commercial production (Burroughs). 
 This speed matches well the anticipated central processor performance using 
 integrated circuit technology. In effect all ILLIAC III circuit and memory 
 speeds would then be increased by a factor of k. An alternate possibility 
 for fast storage, but one without adequate prototype evaluation at the time 
 of writing, is discussed below. 
 
 Certainly one of the more promising fast storage techniques is the 
 plasma memory of Bitzer. Assuming glass plates of 12" x 12" and a hole pattern 
 of 50 bits/linear inch (in both x and y directions), we deduce a storage 
 capacity of 512 x 512 = 262, ikk bits. Assuming a 10-plane module (9 bit 
 information byte + parity), we can envision a quarter million byte memory 
 module of perhaps 18 x 18 x 3 inches, including all electronics. Hypothesized 
 here is single photo multiplier readout using Brewster angle reflection to 
 bring light generated internally in the glass plate sandwich out to the edge. 
 A similar technique was developed by Alvarez for the Scanning/Measuring Projector 
 (SMP). The anticipated low cost of plasma memory suggests that no attempt be 
 made to provide other erasable random access memory. 
 
 One anticipated difficulty with this choice, however, is the unbalance 
 between estimated processor speeds and the fast memory cycle time. Here 
 plasma memory is anticipated to run at approximately 1 usee, cycle time. 
 On the other hand typical arithmetic unit or taxicrinic processor cycle times 
 may be speeded up k-8 times over ILLIAC III performance (because of circuit 
 size, small signal swings and shortening of interconnection wiring). 
 
 This observation suggests 
 
 i) building processors (units) out of MOS technology (at lower unit 
 cost and anticipated better reliability), or 
 ii) using longer "word length" for memory, or 
 
 Professor Donald L. Bitzer, Coordinated Science Laboratory, University of 
 Illinois; private communication. 
 
 -l40- 
 
iii) including at each processor a local MOS buffer memory. 
 
 Objections can be found to all these choices: 
 
 i) The cost advantages of MOS against integrated circuitry do 
 not warrant this choice in a high performance system, 
 ii) The present choice of memory word in ILLIAC III (80 bits) is 
 
 already long. Little programming advantage is contemplated if 
 a unit "word length" for fast memory is selected which 
 exceeds the length of the longest frequently used operand 
 (e.g. floating point number), 
 iii) Use of a local MOS buffer memory requires more programming 
 
 complexity -- particularly in the treatment of interupts. This 
 choice would introduce three levels of memory hierarchy: 32 
 fast registers of the taxicrinic, or input/ output, processor 
 proper (integrated circuit technology), 1024 words of 
 processor local memory (Mps) and finally fast storage (plasma 
 memory) . 
 
 An alternate solution, at this time preferred, would be to make 
 available, possibly independently, to each taxicrinic (or input/output 
 processor a very fast read-only memory. The anticipated objective here is 
 to exploit the novel aspect of ILLIAC III design which permits direct execution 
 of macro instructions. In effect the instruction code which can be interpreted 
 directly without access to fast storage is greatly enriched. This latter 
 choice does not complicate programming nor influence interrupt considerations. 
 Editing and time sharing facilities are in fact enhanced. In line with this 
 possibility the average access rate to fast storage might be reduced by a factor 
 of 2-4, in any case sufficient to place the burden of better processor/fast 
 storage match upon improvements in the plasma memory cycle time. 
 
 6.4 Backup Storage 
 
 Here is one of the truly unpleasant areas of machine design. We 
 assume the storage mechanism employs rotating machinery with worst case 
 access times of 33 msec (l800 rpm) . 
 
 Magnetic disk storage, as presently manufactured, is too costly. 
 One possibility, a compromise between the NCR CRAM Memory and magnetic disk 
 
 -41- 
 
storage, is to use the principle of the a-wrap video tape recorder -- tr< 
 as a single head magnetic drum. For example, the new low cost Ampex vlcU 
 tape recorder stores one hour of recording (108,000 frames) on one roll of 
 magnetic tape. In this unit each 10 inch stripe on the tape is sufficient to 
 make an adequate CRT display as either a broadcast quality TV frame or as 
 approximately twenty lines of digital text. 
 
 This latter type of backup storage might be convenient for scratch 
 pad work. For example in computer-aided design the user may wish to enter 
 working documents of his own origin -- knowing these to contain numerous 
 errors. Here what is needed is a literal scratch pad memory. For this purpose 
 a digitally-indexed unit of the video tape recorder type might be adequate and 
 convenient. Such equipment may also be classified as part of the local input/ 
 output. Documents video taped back at the office with inexpensive equipment 
 could be entered into the machine through a unit of this type. 
 
 For large scale commercial and scientific application, however, we can 
 anticipate that equipment reliability will play a deciding role. Accordingly 
 recording techniques requiring physical contact with the recording surface 
 (e.g. magnetic disks, tapes and drums) are ruled out. We suspect that optical 
 digital storage on photochromic glass disks will dominate the show. Here we 
 suggest photochromic storage of approximately a billion characters (bytes) per 
 disk: 1 bit = O.k \i circumferential, 5 U radial on a 7 inch diameter disk -- 
 approximately 10 bits per revolution, or 30 MC bit rate. 
 
 It is quite possible that photochromic optical disk storage, in 
 conjunction with scratch memory of the video tape recorder type and plasma- 
 memory fast storage, may suffice. Present photochromic memories can be 
 rewritten, but only rather slowly. In general, unlike classical scientific 
 computations such as the numerical solution of partial differential equations, 
 there is a need for extensive backup storage -- but the larger part of this 
 memory does not require fast insertion of individual words. 
 
 For example a requirement for erasable memory arises from the 
 preparation of bibliographic materials: in particular, indexing and 
 cataloging. Too violent reorganization of this material can destroy the service 
 utility of the Information Resource Center. Daily updating can presumably 
 be handled by patches to fast memory. Here microsecond updating of photo- 
 chromic memory is not required. 
 
 -U2- 
 
A similar situation arises with the programming system. The 
 percentage of the entire system changing per day must necessarily be small -- 
 for otherwise older programs of users will not run. We might contemplate a 
 production system where major changes are permitted once per month -- at 
 which time diagnostics must be provided by the system programmer to update 
 older user/ system programs. The use of piiotochromic memory seems entirely 
 reasonable even today for this type operation. 
 
 6.6 Input/ Output 
 
 Input/ output equipment should include standard keyboard 
 telecommunication equipment for the primary input of source programs, cartridge- 
 type digital magnetic tape stations (replacing cards), and other less 
 conventional equipment listed below: 
 
 1) Remote Video Consoles . It is proposed that remote video consoles 
 be established. Each console will include one CRT display 
 screen, one TV camera and a Model 33 ASR teletype set with 
 dataphone connection. These remote video consoles would be 
 dispersed geographically about the site of the Information 
 Resource Center. Also some carrels of the Center would be 
 augmented by a camera for direct visual input. 
 
 2) Microwave Link . One or more remote consoles would be linked 
 to the centralized computer through a video channel carried 
 over a microwave link. This two-way communication over the 
 microwave link will allow us to transact visual data processing 
 at remote sites. 
 
 3) Printer/ Plotter Unit . It is proposed that a xerographic Printer/ 
 Plotter be provided to furnish direct hard copy of alphanumeric, 
 graphic and video information. This printer, when attached to 
 the central computer would allow us to provide immediate 
 documentation for visual image processing (e.g., for microscopy 
 design automation, programming development, etc.) 
 
 h) Magnetic Disk Storage . To provide volatile digital storage 
 essential to realize visual data processing with remote users, 
 "disk-pack" type files will be appended to the central computer. 
 Disk packs provide an interim solution to the backup storage 
 problem, and can be used by the user to load extensive data into 
 
 -^3- 
 
the system. Normally the communications net would be used 
 to input data, thus avoiding the necessity for physical 
 handling of disk packs, magnetic tapes, etc. 
 
 ■kk- 
 
7- PROGRAMMING SYSTEM 
 
 7 -1 Orientation 
 
 Basic to the concept of a Information Resource Center is the 
 idea of a vibrant responsive environment for thought and its immediate 
 effectuation--a widening horizon of our mind's eye . To this end let us 
 speculate about an appropriate strategy to achieve this responsive environ- 
 ment, this enhanced discourse between man and machine. 
 
 Let us review the many transformations in the history of man- 
 machine communication. As a tool of the thinking man, pen and paper, with 
 the ancillary of the publishing and printing industries, have been without 
 peer as instruments of serious communication. Writing derives its 
 magnificent power primarily as a social instrument--as the ideation point 
 to inspire and the controlled environment to coordinate the activities of 
 other individuals in accord with the insights, aims and goals of the author. 
 The "machine" in social history of man-machine communication has been 
 primarily an organization of people: the mill workers, the army, the 
 section crews laying down new railroad track. 
 
 Now contrast this situation with the relatively recent invention 
 called programming. Here again the author communicates through writing-- 
 composing the rather ghoulish character strings of an artificial programming 
 language. The programmer's intention is to cast down a plan (algorithm) 
 for the data acquisition, classification and interpretation of a part of his 
 environment. One difference, however, is central: this sequence of 
 instructions is not to be executed by some remote ensemble of future readers 
 but rather to be executed by an ever-present array of organized electro- 
 mechanical parts--a computer in the broad sense of the term. And with this 
 change has come an intellectual revolution; these orders can be immediately 
 effectuated. This reluctant slave, the computer, can be programmed to 
 interpret our orders, edit our texts, reference our files. It can infuse 
 all these tasks with culturally accumulated pattern recognition and computa- 
 tional skills--and still synchronize with the evolution of our comprehension. 
 
 But beyond those problems well within the scope of contemporary 
 
 -fc5- 
 
programming there remain aspects of man-machine communicatior. are le 
 than ideal- The user still must tediously instruct the machine--at present 
 largely by typing a string of instructions, though graphical technique 
 are also coming in. I surmise that just at this point the game is largely 
 lost. Why for example could not a sufficiently intelligent systems program 
 be written so that the computer could work more directly with the confused 
 incoherent raw materials of our thought? Could not the machine assist 
 in the clarification and expression of thought by providing an anticipatory 
 and responsive environment, visually via the console display and aurally 
 through the headphones? 
 
 Speaking metaphorically, could the computer dream our dreams? 
 The answer here depends critically upon the all-important parameter of privacy. 
 Suppose, for example, the user restricts his computer usage to running 
 Fortran programs. Assuming no personal history of types of problems run, 
 diagnostic errors incurred, etc. is maintained by the machine, his 
 privacy can be very considerable. Suppose, however, the user graduates to 
 a remote console of a computer aided instruction system. In so far as he 
 requests the instruction be sculptured to his needs, it becomes the duty 
 of the teaching program to build up a profile of the user's personality 
 (response-type), maintain a history of test scores, and record concepts found 
 difficult. That is, to increase the responsiveness of this machine-presented 
 environment, the user exposes himself more and sacrifices his privacy further. 
 
 But need we halt here? The Fausts of the world clearly request more- 
 new creative and inventive powers--if at the sacrifice of further privacy. 
 How this responsive environment can be achieved can be illustrated by a simple 
 strategm--the single key typewriter. Let the correspondent of this machine 
 be presented at the console with an initial sequence of characters (letters, 
 digits and punctuation) . The user indicates his choice by pushing the key 
 upon being presented the correct character. Proceeding in this way to the 
 next few characters, the computer begins to anticipate larger units-- 
 tentatively completing common words, e.g. speculatively filling in names, 
 addresses in a letter and retaining these if a touch of the key indicates 
 they should be held. 
 
 By extension this procedure begins to more resemble dictation to 
 an intelligent secretary than to resemble typing. In typing the fingers always 
 
 -1+6- 
 
signal at the same hierarchial level--at the level of the individual 
 character. -With this new keying system the hierarchial level is in 
 anticipation constantly shifting: from the character, to the word, to 
 the paragraph, of the form letter, and back. But we observe, to be 
 effective the single key strategm demands that the user expose more of his 
 mental habits, again at a loss of his privacy. 
 
 Going further, can we take away even this one key? The user 
 willing, we could train a closed circuit TV camera on his eyes and depend 
 entirely upon involuntary communication. By sampling the output of this 
 camera the absolute dilation and transient response of the pupil to material 
 presented at the console gives a measure of user interest—not believed 
 to be subject to conscious control by the participant. (As this is a 
 controlled environment, accurate compensation for changes in overall screen 
 illumination can be instantly made). Further we could monitor the scan 
 pattern as the subject (the term seems more apt than user, at this stage) 
 examines his visual environment. If in rapid-eye -movement (REM) sleep, 
 investigators can occasionally hypothesize the story line of the subject's 
 dream by a study of the sleeper's recorded eye movements, so much more so 
 must it be possible to communicate with those of us who pretend to be awake. 
 
 But before we are further carried away with the use of new psychological 
 probes, it would be well to remark that these new sensory inputs to the 
 machine are just that--new, if non-mechanical, keys for the user to push. 
 What is ever more important is the conceptual structure that permits the 
 augur to transcend from the examination of sheep entrails to the affairs of 
 state. Here a prior careful hierarchial structuring of information is required- 
 extending greatly the simple grammar of character, word, sentence, text for 
 the one-key typewriter. A new facile ability to anticipate new symbols at the 
 same hierarchial level, and more, to transcend or descend in level is required 
 as signalled by the user. 
 
 Systems with this strict attention to hierarchial structure have for 
 years been the basis of many diagnostic routines for computer programs. Under 
 machine control the programmer uses the diagnostic system to descend from 
 statements in a higher level language to the level of individual instructions 
 of the subroutine called. These instructions, in turn, may be imprimitive and 
 
 •hrj- 
 
develop their semantic interpretation only as a nested sequence of 
 
 further subroutine calls. Normally the programmer can examine h. ted 
 
 file of instructions in either of two ways: i) (passively) by scanning 
 
 forwards or backwards the file for content but without instruction 
 
 execution, dropping up or down in hierarchial level as he sees fit, 
 
 and ii) (actively) by serially executing the nested sequence of instructions. 
 
 In this latter mode he serves much as the instruction pointer of the 
 
 machine, indicating when each instruction is to be executed, either in its 
 
 entirety without explicit display of subfiles called by the instruction, or 
 
 partially by entering the defining subfile (if any) . With good console 
 
 display these diagnostic techniques can be made very natural, can be 
 
 extended to apply to data structures as well, and in fact can even incorporate 
 
 + + * ** 
 the anticipatory features of the single key strategm. ' — ' ' 
 
 Now suppose that the user is placed in communication (communion) 
 with an extensive file, say a small library. Suppose the librarian has 
 provided bibliographic materials, and the programmed retrieval algorithms 
 impose this desired, even anticipatory, hierarchial structure. It is likely 
 that the system will fit the user like an old shoe? I conjecture not; 
 the system more likely fits well only the librarian who devised it. Take 
 the word "computer". To me, the word conjures the subtopic descriptors: 
 digital, hybrid, analog, parallel, arithmetic, non-numerical, electronic, 
 mechanical, etc. The librarian, not active in computer research, might 
 have included many other appropriate terms, but have omitted "hybrid". Or 
 the librarian's sublist under "analog" might only weakly correspond to what 
 I have in mind. The system then fails because anticipatory response depends 
 not only on the careful hierarchial structuring of the file, but on the 
 correspondence between subfile listings of the "pool of knowledge" and those 
 of the user . 
 
 An elegant diagnostic system of this kind has been developed by Philip 
 Merryman for the CDC l60Aand more recently, working at the University of 
 Illinois, for the ILLIAC II Computer. 
 
 —For a careful exposition of these concepts see: John A. Wilber, "A 
 Language for Describing Digital Computers", Department of Computer Science 
 Report 197 > University of Illinois, February 1966 . 
 
 * B. H. McCormick and J. A. Wilber, "A Computer-Oriented Language of Computation", 
 Department of Computer Science Report, University of Illinois (in progress) - 
 
 **R. Narasimhan, "Programming Languages and Computers: A Unified Metatheory", 
 Computer Group, Technical Report No. 9, Tata Institute of Fundamental Research, 
 Bombay, March 1Q66 . 
 
 -i+8- 
 
After frequenting the "pool" over a period of months, the user 
 could amend the bibliographic controls of the pool by the addition, 
 replacement and the deletion of items and so build up his own personalized 
 index. Among files so modified might be the anticipation and retrieval 
 instruction listings themselves. Pool entries, whether examined or not, 
 left unmodified, are tentatively assumed valid and retained intact. To 
 the extent to which this process is carried out, the responsiveness of 
 the system to the user's idiom and needs is improved. The file system 
 gains anticipatory responsiveness for this one user. 
 
 But where can the bibliographic materials of the pool come from? 
 Surely it is unreasonable to expect classification procedures that are 
 intellectually significant and current from individuals (editors, librarians, 
 etc.) not active in current research. And here the personal indices provide 
 a natural resolution to this dilemma. We start with an initial pool; 
 examine periodically the personal indices of those men currently at the edge 
 of research; and amend the bibliographic controls of the pool on the basis 
 of this comparison. Accordingly each intelligent user automatically becomes 
 one of the librarians of the collection. We are not unaware of the questions 
 of privacy and copyright involved, but the direction here is obvious--and 
 in fact corresponds loosely to much of the current activity of writing 
 textbooks . 
 
 With files of this degree of structure, programming questions 
 requiring great technical virtuosity arise. what is meant by analogy: e.g. 
 what in the literature of the psychology of learning theory corresponds 
 most directly to the nonparametric training techniques used with linear 
 learning machines of computer science? Two files in different disciplines 
 may exhibit remarkable similarities of format—thereby defining a subset 
 of terms that correspond well. On this basis a selective text in field A 
 might be translated into the terminology of field B. Does a similar text 
 in field B already exist? If so, iterates of this process can be set up 
 with dual translation between fields A and B. The possibilities here are 
 endless . 
 
 The present discussion has been discursive and hardly exhaustive. 
 Rather, all that has been done is to indicate the wealth of intellectual 
 
 Compare J. C R. Licklider, Libraries of the Future, The M.I.T. Press, 
 July 1965, p. 26. Independently the author presented the above concepts, 
 apart from the pupil dilation example, in a course on Computer Graphics, 
 Engineering Summer Conferences, University of Michigan, June 1965- 
 
 -1+9- 
 
ore lying only slightly beneath the range of available information science 
 techniques. And truly as the learning environment is made more responsive 
 to the individual user--the antithesis of mass communication— the techno! -■ 
 is forced to evoke a new epistomology . 
 
 The opportunity here has a close parallel with the development 
 of arithmetic computation during this past decade. Mathematics, prior 
 to the extensive use of computers, gave scant attention to the actual mechanics 
 of computation, e.g. the distinction between an information cell (register) 
 and contents of the cell (value). The development of the programming 
 language Algol 60, perhaps even more than Fortran, first made explicit the 
 types and formats of information required to define an arithmetic computation 
 in a practical and rigorous manner capable of interpretation by machine. 
 These concepts have required a decade for their development and systematic 
 implementation. In like manner files and bibliographic materials date back 
 into antiquity. Yet the specification of a file computation is still in its 
 infancy and the ambiguities of specifying such processes is illuminating 
 how vaguely the concepts of document classification, as also the syntax and 
 semantics of language for fact retrieval, were previously, if at all, defined. 
 
 There are ever-present attempts to set up this new epistemology 
 in complete isolation from the study of individual responses— implying that 
 a preordained "pool of knowledge" exists, and users are imperfect beings 
 whose minds are perverse in so far as this file structure does not fit them. 
 And in part this attitude, if ruthless, is just. We teach one alphabet, 
 one multiplication table, one grammar, one spelling (largely), etc. The 
 socially demanded skills to be inculcated by the grammar school are largely 
 rote execution of simple arithmetic algorithms and the fixed transformations 
 of language from one medium for sensory communication to another. But at 
 a higher educational level this rigid conformity to the dictates of a fixed 
 "pool of knowledge" weakens. And at the level of perhaps greatest interest 
 to the Information Resource Center— the creative and inventive horizon— 
 this concept can prove disastrous. Needed here is the study of classification 
 schemes in action—a comparative anatomy (and physiology) of the processes 
 by which investigators acquire, classify and interpret patterns from the 
 environment of their research. 
 
 ■50- 
 
7 .2 A Grouping of Applications 
 
 Programming for each of the intended applications \see Section 
 2) of an Information Resource Center could of course be developed in 
 complete isolation from all other applications. This strategy however 
 would inhibit the free exchange of techniques common to several applications, 
 and limit the later broadening of the programming base in each applicational 
 area. Alternatively, as proposed here, one can first classify the 
 predominant emphasis of each application into one of five activities and 
 then attempt to maintain programming system design conformity within these 
 subgroupings. 
 
 The five traditional disciplines encompassed by an Information 
 Resource Center are 
 
 1) Data Acquisition, Pattern Recognition and Interpretation : 
 
 laboratory (reactor, accelerator, hospital) automation, 
 industrial process control including pilot plant evaluation; 
 military surveillance such as aerial photo interpretation, 
 satellite weather reconnaissance, and automated radar 
 monitoring; and conventional data processing. 
 
 2) Information Storage and Retrieval : 
 
 library automation, data management systems, medical 
 information control, and professional information exchanges 
 for law, legislative review, etc. 
 
 3) Communications : 
 
 travel reservation and traffic control systems, industrial 
 marketing and distribution networks; nationwide voting 
 and polling systems, and military command and control 
 systems for the field. 
 h) Instruction : 
 
 computer-aided instruction in learning laboratories; 
 audio/visual learning resource centers for the preparation 
 and evaluation of instructional materials; and flight 
 simulators. 
 5) Design Automation : 
 
 computer graphics; system simulation (as modeling of 
 
 -51- 
 
accelerator beam layout, etc); engineering design 
 automation (STRESS, APT, etc); algebraic manipulation 
 and theorem proving; composition and scoring of music; 
 generation and checking of building construction 
 specifications. 
 
 As a mnemonic, with attendant oversimplification, we can say- 
 that these five disciplines emphasize respectively: 
 
 1) Data Acquisition, Pattern Recognition and Interpretation - 
 the recording of new information abstracted from the 
 
 data of one's research environment 
 
 2) Information Storage and Retrieval - the discovery of 
 previously recorded information 
 
 3) Communications - the transmission of information 
 
 h) Instruction - the teaching of information to a user under 
 
 machine domination 
 5) Design Automation - the learning of information processing 
 
 by the machine under human domination. 
 
 The above neat compartmentalization unfortunately leaves out 
 the use of information. For this reason the literature of these disciplines, 
 taken in isolation, often makes for drab reading and claims unsubstantiated 
 by operational experience. The example of junk mail, or computer printouts, 
 amply demonstrates that machines can retrieve and transmit information 
 quite independent of human intervention, and also quite without content. 
 The exponential growth of literature in technology,, science and scholarship 
 now swamps many professional men; the proposal in isolation of an 
 information storage and retrieval system that promises only more abundance 
 elicits little enthusiasm. 
 
 In part, as in the past, the effective use of information can 
 be met only by a strategy of deliberate degradation in the breath of one's 
 area of technical competence and in part by the imaginative transfer to 
 the computer, or supporting staff, of processes previously requiring 
 personal attention. The review function -- keeping one oriented, aware 
 
 I abreast -- will undoubtedly become the domain of the Information Resource 
 Center, at least for those without the financial resources to maintain 
 
 for this purpose. In particular, the center can use its sensitive 
 
 -52- 
 
personal profiles to alleviate the need for indiscriminate junk mailing 
 and the broadcast sowing of technical preprints. Further the programmed 
 comparison of an individual's personal index and reference history 
 with that of the pool can he used to inform the user that his technical 
 competence in a given area has fallen below threshold, and that he might 
 better either request computer-aided instruction or disengage from 
 receiving reports on this area -- to free himself for more productive 
 endeavors elsewhere. In this context, as has often been commented, the 
 computer revolution to date has absorbed many of the more odious tasks 
 (and jobs) of the clerical class, but its contribution to managerial class has 
 been relatively slight. The reason for this situation possibly is as 
 much in the lack of market for the replacement of the latter than any 
 incremental intellectual complexity involved. 
 
 This attention to the use of information, from beginning to end, 
 serves to interrelate the five information processing disciplines. However 
 the concept of the user's information processing needs is elusive. Our 
 problem, I suspect, will be to provide not what the user now requests, 
 but to provide what he should have requested had he better understood the 
 latent powers of the facility. 
 
 By way of illustration, one can refer back to the early days 
 of computers and computer usage. Quite as now bibliographies and citation 
 listings seem to absorb much of the energies of those in the library 
 science field, it was characteristic of the earlier computer workers to use 
 their new machines to compute tables of the Bessel function, etc. As the 
 computer technology advanced, these tables once considered so valuable 
 have been largely abandoned. In fact the real problem facing the user 
 was to solve a numerical computation involving, as one constituent, a 
 Bessel function. To this end he requested a table of the function for 
 a comprehensive range of the argument variable . He would have been better 
 advised to let the computer solve the entire problem, including the 
 computation of the Bessel function for the specific values of the variable 
 required. Instead he was diverted into the largely irrelevant task of 
 computing these tables -- the right solution to the wrong problem. 
 
 With this historical hindsight, the programming system design 
 of an Information Resource Center should proceed, not from a preconceived 
 
 ■53- 
 
list of fixed services to be provided, but from the announced class of 
 problems the user really wants solved. 
 
 -5h- 
 
8- COMMUNICATION NET 
 
 8 .1 General 
 
 The Communication Net of the Information Resource Center will 
 provide remote users with data and control communication to the central 
 computer complex, the file and display subsystem of the center, and under 
 conditions detailed below, to other users of the net. 
 
 The Communication Net will allow three distinct modalities of 
 information communication: 
 
 1) dataphone connection , over Bell Telephone lines 
 
 2) high speed digital character transmission , over 36 parallel 
 coaxial cables, and 
 
 3) video communication , over low dispersion closed circuit 
 television coaxial cables. 
 
 It is our intention that the dataphone connection serve remote 
 Teletype sets and low data rate experimental equipment. Commercially purchased 
 laboratory equipment to be used with the Communication Net should be provided 
 by the manufacturer with direct dataphone output--and, where convenient, the 
 added option of being able to drive a monitoring Teletype set. 
 
 There are however high data rate experimental equipments, scan/ 
 display systems, data acquisition systems and remote general purpose computers 
 for which the channel capacity of telephone lines is prohibitively low. For 
 these subsystems the Communication Net will provide direct coaxial communication 
 over 36 parallel coaxial cables. The digital interface, as seen by the user, 
 is an extension of the standard IBM System/ 36O I/O Interface (abbreviated 
 henceforth as X36O interface ) . Our inquiries have shown that most equipment 
 can now be purchased in this interface--whether the equipment is of IBM 
 or other manufacture. 
 
 Finally, to exploit the unique visual data processing capabilities 
 of the central computer, in particular the pattern articulation units, a 
 video modality will be appended to the Communication Net. The intention here 
 is to allow closed circuit television cameras, monitors, video tape recorders, 
 etc. to be used as remote video consoles of the Information Resource Center. 
 
 -55- 
 
Specifically this network comprises two (2) video signal lines (CCTV 
 coaxial cable), three (3) digital control/ synchronization lines (coaxial 
 cable) and various standard camera/ monitor remote control lines (twisted 
 pair) . 
 
 8.2 Applications 
 
 The transmission of information is an activity in of itself 
 having no great intellectual content, and contributes to the economy only 
 insofar as the flow of bits, as of water, electricity or any other commodity 
 transported by a public utility, is used sensibly at its terminals. 
 
 The proposal of a long distance communication net opens the question 
 of what services should be provided locally by an Information Resource Center 
 and what services should be relegated to some state, national or even 
 international hierarchy of stations. We observe first that the Information 
 Resource Center has been conceived here as a largely specialized institution- 
 appendage d here to a city hospital, there to a biomedical laboratory, to 
 an accelerator site, etc. Accordingly the users most likely to be in need 
 of intimate communication are predominately local to that center, and the 
 need for extensive detailed communication (other than by the publishing of 
 programs, data, technical reports, etc) attenuates sharply with distance. 
 The literature required by this group quite likely can be found within the 
 100,000 book (or equivalent record) files of the local center. And though 
 microwave links are proposed between neighboring centers, local autonomy 
 is not seriously jeopardized here as the microwave communication typically 
 would require less than five percent of the channel capacity of the center. 
 
 Therefore the immediate need for a small number of large scale 
 centralized facilities is seen to be largely specious. Rather of much 
 greater importance is the exuberant growth of local centers, individually 
 dedicated to their specialized tasks, such that an indigenous population of 
 users, programmers, and techniques (theoretical and applied) can develop 
 independently of arbitrary bureaucratic constraints and premature standard- 
 ization. 
 
 In accord with our attention to local initiative, our interest in 
 ;he user is in proportion to the bandwidth (channel capacity) afforded him: 
 ;imes more for the user within the Information Resource Center proper, 
 
 -56- 
 
10 times more for the user 1000 yards away, than to the attention granted, the user 
 at some remote distance linked only by telephone. This situation of course 
 only reflects the economics of any communication net. Long distance 
 xerography, for example, does not approach the quality nor economy of 
 commercial printing. One pays for bandwidth, and at the proposed cost per 
 installation the balance for the next decade will be predominately in favor 
 of many local centers over centralized distribution. 
 
 8-2.1 Typical Telecommunication Applications 
 
 hospital ward monitoring and remote recording of EKG, etc. 
 
 data acquisition from low data rate experimental equipment 
 
 data acquisition on invoices, sales, shipping records, etc. 
 
 ticket reservation systems 
 
 merchandizing mart/ commodities exchange automation 
 
 industrial process control 
 
 machine tool digital control and digital plotting 
 
 type setting/music score matrix generation 
 
 automatic bibliographic reference systems for libraries 
 
 long distance xerographic document transmission 
 
 change order information for 
 
 teaching machines (with local text storage), 
 
 remote drafting consoles (with local drum storage), and 
 
 CRT storage/ display consoles 
 
 8.2.2 Typical High Speed Character Transmission Applications 
 
 computer -to- computer communication 
 
 film and microscope slide scanners 
 
 data acquisition from high data rate experimental equipment 
 
 industrial quality control (microcircuit testing, etc.) 
 
 alphanumeric encoding of printed text 
 
 special scan applications: ultrasound scan of viscera, 
 
 restricted scans for pathological cells, radar surveillance. 
 
 8.2-3 Typical Video Communication Applications 
 
 advisor communication (for assistance with system programs, etc.) 
 
 ■57- 
 
visual data processing from remote stations 
 
 preparation of instructional materials (audio/visual aids, 
 
 teaching machine scripts, film, etc.) 
 graphic arts automation 
 microform reference library backup for junior colleges, 
 
 technical institutes, underdeveloped countries, etc. 
 regional backup libraries using microform texts 
 coupling to educational TV and CATV networks 
 computer-aided instruction 
 satellite weather reconnaissance 
 data management systems requiring visual display 
 
 8 = 3 Trunk Lines 
 
 We distinguish between external trunk lines of the communication net, 
 which permit remote users direct multi-coaxial cable connection to the 
 central processors of the Information Resource Center, and internal trunk 
 lines, which service input/output equipment and the file/ display subsystem 
 housed directly in the Information Resource Center. The computer system 
 does not functionally distinguish between external and internal lines, 
 except through program-controlled priority-of-service constraints. 
 
 Each trunk line, external or internal, is assigned one I/O 
 channel, and interfaces with the central processors through a separate 
 Channel Interface Unit. In the prototype ILLIAC III computer system five 
 channels are associated with five external trunk line respectively, and 
 eleven channels are associated with the internal trunks. These sixteen 
 trunks saturate the full sixteen channel capacity of the computer. However 
 it is anticipated that this number of channels is minimal, and for a 
 Information Resource Center with an extensive communications net 32 or even 
 6U channels will be more commonly the case. 
 
 8-*+ Teletypewriter Equipment for Dataphone Connection 
 
 The Communication Net will allow, as one of the three distinct 
 modalities of information interchange, direct dataphone connection over 
 Telephone lines. Aside from the immediate transfer of experimental 
 
 .ow transmission rates, the predominant use of dataphone connections 
 
 -58- 
 
to the Information Resource Center will be to service Teletype remote 
 consoles . 
 
 All consoles will employ a subset of the American Standard Code 
 for Information Interchange (ASCII). This code is used uniformly with all 
 I/O devices of the center. Presently the preferred consoles are essentially 
 Model ASR33 Teletype Automatic Send-Receive Sets in the "Switched Network 
 Version". With recommended modifications these sets incorporate the 
 "Answer-Back", "Pin-Feed" and "Form Out" features. But of greater technical 
 importance the sets purchased direct from Teletype Corporation are provided 
 with parity generation. 
 
 8 • 5 Description of High Speed Digital Character Transmission Interface 
 
 High speed information transmission, parallel by character, between 
 Information Resource Center and remote attached devices is governed in format 
 of busses, mode of operation (device addressing, commands, sequence controls, 
 etc.) and physical signal levels as specified by the conventions of the 
 IBM System/360 I/O Interface -- as extended below. Description of this 
 interface will be found in literature supplied by IBM to original equipment 
 manufactures as File No. S/36O-I9, Form A22-68 1 43- 
 
 This interface is extended (hence X36O) in the computer system of 
 the Information Resource Center in two regards: 
 
 l) Nine bit character transmission is permitted. Accordingly 
 the INBUS and the OUTBUS are each augmented by an 
 additional bus line 8. Accordingly a (10 bit) bus character 
 looks like: 
 Parity . . „-- - 
 
 fla* 
 
 0123^5678 
 < data byte 3 
 
 As before the byte has odd parity. For devices 
 
 not employing the flag bit in each byte (i.e. IBM System/ 
 
 360 equipment), line 8 can be left as an open input (cable 
 
 logical zero) . 
 
 In the normal IBM 36O interface all transmission is interlocked 
 
 -59- 
 
with corresponding response signals, permitting system 
 input and output operations that are not dependent upc 
 circuit speed. However, this strict reply-back arrangement 
 can restrict data transmission to approximately 10% of 
 maximum channel capacity -- because of the propagation delays 
 of these replies in the extensive 36-line coaxial cable 
 trunk system. Accordingly we will permit stream transmission , 
 i.e. transmission of a prespecified number of data bytes 
 at a prearranged clocked rate. Dummy replies will be 
 generated by the communication net -- so that as seen by 
 the attached device the transmission is, apart from the 
 prearranged byte transmission frequency, indistinguishable 
 from a local channel connection operating in burst mode. 
 Only (multibyte) data transmission can operate in this 
 stream transmission mode. Initial device selection and the 
 ending procedure are fully interlocked with reply-back 
 signals. 
 
 Maximum character rates for transmission depend of course upon 
 the distance from the Information Resource Center. Design specification 
 of 5 MC character rate for distances under 500 meters, and 1 MC character 
 rate for distances up to 3000 meters for coaxial cable transmission appears 
 reasonable. Assuming digital encoding these rates correspond to l.k seconds 
 for the transmission of a typical textbook over a radius of approximately 
 two miles, and 0.28 seconds for more local transmission. Alternatively the 
 same material could be sent by high resolution closed circuit television (see 
 Sec. 8-8) using the 8 MC bandwidth video communication net. At 7 . 5 
 frames/ sec the corresponding transmission time would be 67 seconds by this option; 
 however virtually no computer time would be required to reconstruct the text 
 and figures. 
 
 8.6 Description of Video Interface 
 
 Video transmission between the Information Resource Center as well 
 as between one video console and another, is governed in format of lines, 
 
 of operation (device addressing, commands, sequence controls, etc.) and 
 .cal signal levels as specified by the conventions of the video interface 
 specified be] 
 
 -60- 
 
These video interface standards can be understood by analogy 
 with the X36O interface conventions. The video cable contains an OUTBUS 
 (a low dispersion coaxial cable for the outgoing video signal + twisted 
 pair lines for outgoing camera/ monitor remote control signals), an INBUS 
 (a low dispersion coaxial cable for the incoming video signal + twisted 
 pair lines for incoming camera/monitor remote control signals), a single 
 outgoing control line OUTCNTL (digital-quality coaxial cable), a single 
 control line INCNTL (digital-quality coaxial cable), and a synchronization 
 line SYNC (digital-quality coaxial cable). Basically the control lines, 
 OUTCNTL and INCNTL, mimic in bit serial format the multiple control lines of 
 the X36O interface. 
 
 The video interface assumes that time slots of l/30th second are 
 assigned a given video console by the communication net supervisory program. 
 Initial device selection and the ending procedure are permitted only at the 
 initiation and termination respectively of a time slot and employ the INCNTL 
 and OUTCNTL lines exclusively. Accordingly the video interface assumes a 
 data transmission sequence (over INBUS or OUTBUS lines) which is an integral 
 multiple of l/30 sec (approximately) . This is the EIA-specif ied frame 
 time for 525-line broadcast TV. 
 
 Approximately ik twisted pair conductors are assigned the low 
 frequency camera/monitor control functions: focus, pan, tilt, zoom lens 
 control, etc. The detailed assignment of these lines will follow commercial 
 television practice. 
 
 -6l- 
 
9- AUTOMATED LABORATORY 
 
 9.1 General 
 
 "Automated Laboratory" as applied in this note will refer to a 
 research laboratory employing the macro-module concept of instrumentation 
 within a building specifically designed to give this concept maximum sway. 
 It will be implied that the laboratory couples to a Information Resource 
 
 Center. nearby. 
 
 + 
 The macro-modular instrumentation concept -- espoused most 
 
 vigorously by the Computer Research Laboratory, Washington University, St. 
 
 Louis, -- is a bold program to provide the institutional research laboratories, 
 
 university or other., with experimental process control and data acquisition 
 
 facilities previously only economically possible for segments of the chemical 
 
 industry. The "Macro -module" system refers to a family of compatible 
 
 computer subassemblies and electrical, physical and chemical transducers 
 
 which can be linked together and used by an experimenter not trained in computer 
 
 engineering. For example, a "cable" in this system is not merely a harness 
 
 of coaxial conductors with terminal connectors, but includes a selfpowered 
 
 logical assembly to provide the sequence of interface signals essential to 
 
 address and reliably transmit information within a multi-computer complex. 
 
 Modular design is important first to give maximum flexibility for building 
 
 up new experimental complexes, and secondly to minimize the inventory of 
 
 modules required to implement a given wide range of experiments. 
 
 9-2 Architectural Over View of an Automated Laboratory 
 
 A first picture would project the automated laboratory as a 
 two-story free span building perhaps 80' x 120' . Around the periphery of 
 the building would be a regular sequence of precast concrete structural 
 members. These members, or struts, follow the general description of the 
 console-type concrete shell -- hollow U-shaped members 8' wide, 6' deep 
 and2V high. Narrow floor-to-ceiling slit windows separate these concrete 
 
 . segments. As these struts carry little structural weight, they might 
 be precast quite thin with an insulation filler. 
 
 cki and Ornstein, "A Macromodular Approach to Computer Design", 
 :hnical Report No. 1, Computer Research Laboratory, Washington University, 
 >66 _6 2 _ 
 
120' 
 
 2'H h-8'H 
 
 iniririririnririnn. 
 
 Figure 9* Tentative Floor -Plan for Automated Laboratory 
 
 The first floor interior of each concrete strut serves as a "pod" 
 to hold any of many different (and hence compatible) standard packages. 
 Possible standard modules could include: 
 
 2.1 Console and desk 
 
 2.2 Conventional laboratory bench with drawers 
 2 . 3 Fume hood 
 
 2.k Polyethylene acid sink with work area 
 
 2.5 Local clean area with laminar flow benches 
 
 2.6 Macro-module electronics rack with multiplexor connections 
 
 2.7 Dishwasher and autoclave 
 
 2.8 Magnetic/acoustic isolation booth 
 2-9 Darkroom facility 
 
 2.10 Vacuum system: vac -ion or diffusion 
 
 2.11 Animal handling cages 
 
 2.12 Environment control unit to provide special air handling 
 (humidity, temperature, noxious gas control, etc.) 
 
 It is contemplated that all standard packages fit identical 
 electrical and plumbing connections present in every concrete pod. In 
 
 -63- 
 
particular, precast into every strut is transite ductwork so that fumehood 
 connection is always possible. (Accordingly a fumehood module might 
 include not only the blower facility to exhaust fumes, but also an air 
 handling unit with outside intake to make up exhausted air. Alternatively 
 a computed load of ceiling mounted air conditioners can be installed.) 
 These conventionally used modules would be designed where possible around 
 commercially available laboratory equipment. Installation would require 
 neither an electrician nor plumber, and in fact could be installed, or 
 disconnected, in 10 minutes by an untrained technician. These pod 
 "packages" would be fitted with retractable casters for rolling in place, 
 and with lifting eye for ready transportation by crane. 
 
 The floor of the experimental area would be crisscrossed with 
 covered trenches for laying cable and providing take-offs for electrical 
 power, water, air, gas, etc. and interface lines to the communication net. The 
 experience of accelerator experiment areas at Argonne, Brookhaven, etc. 
 could provide prototype specifications and recommended safety precautions. 
 
 Experimental modules of medium size and custom design would be 
 built to standardized base constraints specifying tap positions, connector 
 types and rating for electricity, water, signal cables, etc. A proposed 
 30" x 30" base plate convention might be uniformly used, with bases of 
 60" x 30", 90" x 30", etc. permitted only when dictated by mechanical 
 layout problems. In this spirit we might think of this standardized module 
 base as the analogue of a male printed circuit connector; the experimental 
 area, a giant checkerboard regularly partitioned off into 30" x 30" cells, 
 as the equivalent of the female connectors of a computer backboard. 
 
 It is entirely obvious that the utilities pattern must mesh with the 
 rows and columns of base cells. Raceways on 60" centers might be suggested. 
 A utilities pattern of this complexity can be economically cast into concrete 
 only if extensive prefabrication to a standardized pattern is worked out -- 
 i.e. another architectural "module" of the building. 
 
 We can now view the entire experimental area like an opera theatre -- 
 with drop walls and unused but current experimental modules assembled like so 
 
 current sets over unused areas of the experimental floor. Less frequently 
 I experimental equipment, or standard pod modules shared by several 
 
 -61+- 
 
laboratories, would be removed to a remote warehouse area for inexpensive 
 storage or restoration -- quite as in opera company practice. 
 
 9-3 Use of the Experimental Area 
 
 In general an investigator using the Automated Laboratory is 
 assigned floor space on the basis of a review of the proposed experiments 
 (scheduling committee). He may, for example, require space for only 6 weeks/ 
 years -- using remaining time to interpret his data, prepare publications 
 and discharge other unrelated duties. Fabrication and initial checkout 
 of experimental equipment is of course relegated to lower cost warehouse/ shop 
 area -- until such time as the conveniences and automatic data acquisition 
 features of the experimental area are warranted. 
 
 The investigator, once granted floor space, orders installed the 
 standard pod modules appropriate to this experiment: fume hoods, laboratory 
 benches, vacuum equipment, etc. His custom experimental equipment, prepared 
 by a subcontractor or detached machine shop, has been preassembled on the 
 standard 30" x 30" bases. Within hours his area is prepared: pods filled, 
 custom or previously assembled modules lifted in by crane and installed, 
 computer-type instrumentation macro-modules lashed up, partitioning walls, 
 overhead lighting and air conditioning panels (similar to units designed by 
 the School Construction Systems Development; dropped in place. At this 
 time both the experiment and the experimenter are on-line both to the 
 Information Resource Center and to the locally-controlled macro-module 
 instrumentation/ computation net. 
 
 We note certain changes of practice here. First the unit cost of 
 custom facilities is shared over as large a number of experiments as possible. 
 Office space, in general, would be nearby but not in this building proper. 
 Experimental control, data acquisition and interpretation is in part remoted 
 through the communications net to areas better adapted to these tasks. Finally 
 the individual experimenter has access to professional support both for 
 instrumentation for data acquisition and programming for information processing. 
 These support groups, serving a wider community and challenged by the potentials 
 
 + School Planning Laboratory Reports, School of Education, Stanford University, 
 Stanford, California. 
 
 -65- 
 
of the system, could be expected to be of superior quality to individuals 
 attracted on a piecemeal basis. 
 
 If the range of contemplated experimental equipment is sufficiently 
 light weight, a high-rise building can be used and the overhead crane 
 deleted. Pod packages are then wheeled into place, and wall rearrangements 
 could follow the conventions of the Stanford School Construction System 
 Development Project. Another variant has a crane per floor, allotting 2-3 
 feet for this purpose. These alternatives however do not approach the 
 freedom of the first design with an entire story allotted the crane for 
 transportation . 
 
 It is to be recognized that schema approaching this degree of 
 modularity have already been tried. For example experimental areas around 
 high energy accelerators use a regular grid of utilities, presently 
 standardize equipment design where feasible, and assign time by experiment. 
 Nuclear counter groups at these installations now conventionally use 
 standardized counter modules, signed out from a common pool for each experiment. 
 These counter modules are then maintained and technologically updated by a 
 professional group assigned this sole task. Also of course the treatment of 
 the experimental area follows current practice in some of the more advanced 
 production and warehousing companies. 
 
 A statistical study of contract scientific and engineering research 
 and research equipment at the University could help evaluate design parameters 
 for this type laboratory. Here modularization of space, utility requirements, 
 flow of people, characteristic instrumentation and data rates, sources of 
 information, etc. could be classified as an aid to design. 
 
 9-h The Small Automated Laboratory 
 
 It is anticipated that the standard pod modules could include all 
 packages required for a small independent, though automated, laboratory -- 
 e.g. for clinical pathology. Here the advantages of modular construction center 
 upon high volume prefabricated installation. The proposed design abates in part the 
 ever present problem of modernization, without building remodeling, in an 
 era of rapid advances in bio-medical instrumentation. 
 
 -66- 
 
10= SUMMARY 
 
 10.1 Restatement of Goals 
 
 As seen by the user, the goals of an Information Resource Center 
 are both personal and applicational. To the individual user we postulated 
 a new experimental environment: 
 
 "a setting with information sources to saturate his ability 
 to learn, to formulate new questions, and to record his 
 responses" . 
 
 The potential application of this user environment was delineated 
 for a limited number of areas ideally suited to the proposed techniques: 
 
 1) Medical Information Centers for the programmed interaction 
 of hospitals (medical records, patient monitoring, billing, 
 etc.); doctors (disease diagnosis, correlation of laboratory 
 tests, summary of patient medical records, billing, etc.), 
 and patients (taking of medical histories, mental inventory, 
 etc.). Similar techniques should be extended into the area 
 of public mental health. 
 
 2) Data Acquisition and Interpretation Centers for visual data 
 processing in high energy physics (bubble chamber and spark 
 chamber particle detection techniques), in biology (light 
 and electron microscopy), or in meterology (weather satellite 
 data acquisition and analysis). Analogous techniques are 
 applicable to aerial photo interpretation. 
 
 3) Extension to the Automated Laboratory , e.g. for the synthesis 
 of genetic materials beyond the expectation of an 
 exclusively human coordinated experiment. In like manner 
 
 the potentialities for brain research--an automated laboratory 
 with attached facilities for electroneural probing, ultrasonic 
 lesioning, and the automatic scanning of neural histological 
 materials—are largely unexplored. Here the local microform library 
 keeps reference materials in control and contains the various 
 brain atlases, 
 k) An experimental vehicle for Library Automation , suggesting in 
 
 -6 T - 
 
time the resuscitation of the conventional research library. 
 In particular the law library, where great indexing in depth 
 is required, may be provided new conceptual transparency 
 this approach. 
 5) Training and Assistance of the Less Skilled Individual 
 
 Several instances of employing individuals "on-line" to a computer 
 to economically augment the cognitive powers of the computer 
 complex are now known. The extension to nursing, teaching of 
 culturally deprived children, and auto repair of these "on-line" 
 techniques can be considered a prolongation of present computer- 
 aided instruction—the less skilled individual is provided 
 guidance and ancillary information to allow him to compete as 
 a skilled technician. 
 
 It is a central thesis of this report that though these goals will be 
 separately attached, they will notbe separately attained. We have postulated 
 a central core--the proper concern of the computer sciences—transferable from 
 one applicational to another and accounting perhaps for at least 50$ of the 
 programming and design effort required. 
 
 10.2 Feasibility 
 
 The hardware realization of the Information Resource Center entails 
 development of the building, display consoles, microform file, the computer 
 (central processors, fast storage and bulk storage), and local input/ output 
 equipment. In each instance this report has established either the existence 
 of prototype equipment, which could be suitably modified, or demonstrated that 
 parallel developments (largely in industry) can be expected to implement the 
 demands developed above. Accordingly centers of this type are well within the 
 projected state of the art these next few years. 
 
 The software realization of the Information Resource Center must of 
 course be attacked in stages. Some areas could be transplated now: contemporary 
 bubble chamber data analysis, or a restricted form of hospital information system 
 with nursing stations, connections to the pharmacy, etc. 
 
 As a second level of sophistication the data bases of these areas 
 needs to be reexamined and common (i.e. application-independent) file techniques 
 applied uniformly. Man-machine interface at this stage could be improved by 
 
 -68- 
 
the introduction of English macro grammars. 
 
 And as experience and programming tools permit, we could attack the 
 basic concept of the "mind's eye", wherein the careful hierarchial structuring 
 of information allows the user to readily communicate at the highest syntactic 
 level he can successfully signal across the man-machine interface. 
 
 10-3 Economic Attractiveness 
 
 The concept of an Information Resource Center as developed here is 
 largely a response to the obvious needs of the institutional world (universities 
 and junior colleges, research libraries, hospitals and mental health institutions) 
 Not-for-profit corporations of this type reorganize their informational 
 resources in response to economic pressures only very slowly. To aid this world 
 the cost of a center must not exceed approximately 1 million dollars. A higher 
 price per installation has many deleterious effects: 
 
 1) it delays acceptance of the new technology, 
 
 2) it accentuates the ascendancy of the best (or wealthiest) 
 institutions—as these alone can acquire the new equipment, 
 and 
 
 3) it fosters political empire building, forcing less wealthy 
 institutions to band together under the tutelage of another 
 politically oriented institution. Much of the current 
 interest in inter-institutional communication via computer, 
 etc. largely reflects these political aspirations and only 
 nominally the attainment of new local proficiency. 
 
 Accordingly attention has been taken to strip the Information 
 Resource Center of facilities that only marginally increase its utility. 
 However two commonly suggested modes of economy have been strongly resisted: 
 
 i) decreased computational power of the central computer 
 
 (of marginal economic advantage and extremely deleterious 
 to the attainable level of man-machine communication) and 
 ii) crowding of the consoles into a "locker room" arrangement-- 
 solemn rows of identical consoles like the conventional 
 arrangement of a language laboratory. There is little 
 economic, and no aesthetic, justification for this dull 
 marraige of man and machine. It is not uncommon for a 
 
computer center to enshrine the machine (computer) in a glass 
 palace and then push the users (programmers) into steamy cubicles 
 reminiscent of quarters on a slave ship. The proposal here 
 reverses this trend; the computer is treated on a footing 
 with the mechanical plant. The computer hardware complex 
 should be of no more direct inportance to the user than 
 whether hot air or a hot water ventilation system is used 
 within the building. 
 
 The fundamental premise here is that the cost of an Information 
 Resource Center depends most critically upon volume of the market. The more 
 variation in equipment between applications, the less equipment any one 
 installation will be able to afford. 
 
 Finally the design here has been selected to be modular. With 
 nominal foresight the initial building can be treated as a module of a larger 
 complex--with connecting passageways if nearby, or with interconnecting coaxial 
 cable wireways if more remote connections are desired. Therefore natural 
 growth can be accommodated. The problem of signal interconnection between 
 building modules is not different in kind, and in fact can be considered a 
 special case, of the problems of setting up a communications net (Chapter 8). 
 
 10. h Stimulus to the Cognitive Sciences 
 
 The study of cognitive systems is currently largely the study of 
 biological systems. For pedagogical purposes the life sciences are increasingly 
 taught as the separate study of cells, organisms and populations. This 
 latter subdivision might well be also adopted by the nascent computer sciences: 
 
 i) cells -»■ microelectronics 
 ii) organisms -» computers and associated input/ output equipment, and 
 iii) populations -* problems of programming languages, time-sharing, 
 libraries for recorded intellectual history, information 
 resource centers and their networks. 
 
 The history of the computer development of the past twenty years shows a 
 steady progression in the relative attention paid (first) to the cell, then 
 to the organism, and most recently to the population. 
 
 The time is now ripe to proceed to the population level--and in 
 
 -70- 
 
particular to extend our knowledge of computer science. One sees here a 
 three-pronged investigation--f irst the psychology of the man-machine 
 interface; secondly the better implementation of language processing and 
 other human cognitive skills within the machine; and finally an attack upon 
 the parallel problem of information processing within the central nervous 
 system itself. The order here clearly reflects the relative time scale 
 for maturation of these investigations. It is not unlikely that advances 
 in areas 1, 2 will provide the tools for the analysis of area 3- The 
 illustration of sight-sensory systems however provides a quiet rejoinder. 
 It was the neurophysiological investigation of local feature extraction that 
 suggested first, several hardware realizations of visual pattern recognition, 
 and secondly, has provided psychology with the first non-trivial insight into 
 the visual perception. Now the neurphysiologist .is beginning to examine the 
 utilization of visual information to move muscles, etc . --information to what avail 
 
 Conclusion : the Information Resource Center provides a convenient 
 vehicle for a necessary (population oriented) phase in the development of 
 computer science. 
 
 -71- 
 
APPENDIX A: SUBCONTRACTING AND COST CONSIDERATIONS 
 
 A.l Principal Constituents 
 
 For the Information Resource Center we can assign areas of 
 development as follows: 
 
 1. Modular Information Resource Center Building 
 
 2. Architectural Studies for Associated Facilities 
 3- Display Consoles 
 
 k. Microform Files 
 
 5. Computer Central Processors 
 
 6. Fast Storage 
 
 7. Backup Storage 
 
 8. Local Input/ Output 
 
 9- Communication Net to Remote Users 
 
 10. Programming System 
 
 11. Applicational Programming 
 
 12. Document Processing System 
 
 For an Automated Laboratory, considered as an integral extension 
 of the above development, we must develop in addition: 
 
 13' Modular Automated Laboratory building 
 
 Ik. At chitectural Studies for Associated Facilities 
 
 15. Standard Laboratory Pod Packages 
 
 16. Computer Macro-Modules 
 17- Transducer Macro-Modules 
 
 18 • Extended Programming System 
 
 19' Extended Applicational Programming 
 
 A. 2 Target Production Costs 
 
 One consideration, cost, per installation, will dominate the 
 acceptance and adoption of Information Resource Centers by the institutional 
 world: the universities, junior colleges, hospitals, mental health clinics, 
 etc. Accordingly commercial interest in an Information Resource system 
 of this sophistication depends highly upon production-type cost figures. 
 
 -72- 
 
Installation costs summarized below do not include programming costs and 
 document processing services (which could be distributed over many centers), 
 nor do they include development costs. 
 
 We assume here a two story building with basement, approximately 
 1^,000 sq. ft. in area, containing 50 consoles of assorted types and 
 6k modules (100 disks each) of the microform disk file (as illustrated in 
 Section k) . We estimate, apart from development costs, as follows: 
 
 2.1 Building 
 
 2.2 Display Consoles 
 
 2.3 Microform Files 
 2.U Computer: Central 
 
 Processors and Fast 
 Storage 
 
 2.5 Bulk Storage 
 
 2.6 Input/ Output (Local. 
 
 with known 
 (non-optimal) 
 solution 
 
 with product 
 development 
 
 (Fall 68) 
 $300,000 
 
 200,000 
 
 350,000 
 
 300,000 
 
 200,000 
 150,000 
 
 $1,500,000 
 
 (Spring 70) 
 $250,000 
 
 100,000 
 
 250,000 
 
 100,000 
 
 50,000 
 50,000 
 
 $800,000 
 
 The cost of the Communications Net would be charged to remote users and 
 hence has not been added to the above estimates. 
 
 If the entire cost of the installation were amortized in 5 years and 
 charged entirely to local users, then- 
 
 Cost per Console (life) 
 Cost per Console (per hour) 
 
 Fall 1 
 
 $30,000 
 $ +3.00 
 
 Spring 1970 
 
 $16,000 
 $ 1.60 
 
 To these costs must be charged operating and maintenance costs, photographic 
 services and printing costs, very considerable programming costs, and return 
 on the investment. On the other side of the ledger one can deduct backup 
 data processing carried out on the other 3 shifts/week, and charges to the 
 
 Assuming 2000 hours of operation/ year and five year equipment 
 
 -73- 
 
remote users of the communications net -- e.g. to an attached automated laboratory 
 
 These hardware costs are approaching the level where a franchise-type 
 corporation (similar to Howard Johnson, Holiday Inns, or AMF/Brunswick 
 Bowling Lanes) could seriously contemplate operation. Computer subsystems, 
 though obviously more sophisticated, are not different in principle from 
 pin setting equipment, and the managerial skills to run an existing 
 information resource center are probably not fundamentally different from 
 those to run a bowling lane franchise. And assuredly the microform disk 
 file, designed around the commercial juke box record changer, could be 
 updated and maintained by a separate service organization with obvious 
 parallels (if unsavory) in the older re cord/ vending machine industry. 
 
 Software costs however are a very different matter. At the 
 time of writing no dependable estimate is known for the programming cost 
 to achieve installations with the full generality described in the text. 
 Per applicational area, 1,000,000 instructions is certainly minimal, and 
 even at academic rates ($V instruction, including planning) this comes 
 to $U, 000, 000/ area. Presumably most conventional system programming, file 
 processing routines, pattern recognition algorithms, etc. can be transplanted 
 (as discussed in Section 7*2) to other applicational areas at reduced cost. 
 But a conservative estimate would still place programming costs proportional 
 to the number of independent applicational areas. 
 
 It is unlikely, I believe, that a university, a not-for-profit 
 corporation set up specifically for this development, or a profit corporation 
 alone will have enviable success in this type endeavor. The use of natural 
 language data processing, pattern recognition, techniques, fact interrogation 
 of file structures, etc. are all theoretical topics at the core of a 
 university curriculum in information sciences. An adequate base for the 
 programming, however, requires that the field be rapidly populated with 
 compatibly similar centers -- in other words, that a normal commercial 
 market be readily developed. And the not-for-profit corporation, sandwiched 
 between the other two, often assumes the worst attributes of both. 
 
 A- 3 Subsystem/ Subgroup Responsibilities 
 
 The developmental and programming tasks to realize an operational 
 ormation resource center can be classified by grouping together those 
 
 - T U- 
 
tasks requiring similar professional training and technology (e.g. computer 
 engineering, library science, programming, etc.), or alternately by 
 combining together tasks related to a common subsystem of the center. 
 Here largely the latter description has been chosen; the information 
 resource center, with option of an attached automated laboratory, 
 has been partitioned into subsystems which in principle could be tested to 
 independent performance specifications. With each subsystem area is 
 associated a coordinating group with responsibilities outlined below. 
 
 A. 3-1 Architectural and Industrial Design Group 
 
 Responsibilities of this group comprise: 
 
 1.1 Information Resource Center Building (design; interior 
 decoration; mechanical, plumbing, ventilation, air 
 conditioning, and electrical work including cable 
 raceways, wiring harnesses, etc.) 
 
 1.2 Design and Construction of Associated Facilities (office 
 space, etc . ) 
 
 1-3 Carrel Design and Display Console Packaging 
 l.k Microform File Packaging 
 
 1.5 Computer Packaging 
 
 If an attached automated laboratory is planned, then the 
 responsibilities of this group might be enlarged to include: 
 
 1.6 Automated Laboratory Building (including cables of the 
 communications net) 
 
 1.7 Design and Construction of Associated Facilities (shop 
 areas, etc . ) 
 
 1.8 Laboratory Pod Packages 
 
 1.9 Base Modules and Standards for Larger Experimental Equipment 
 
 1.10 Macro-Module Packaging 
 
 The intention here is to place responsibility in one group for 
 the physical configuration, human engineering and visual aesthetics of 
 the entire system. Development of the "carrel wall" with display console 
 inserts will require a merger of architectural and industrial design. In 
 like manner the packaging of macro-modules must reflect a merger of 
 
 -75- 
 
industrial design and computer-engineering specified constraints (unit 
 
 fan out, power ratings, etc) such that what can be fitted together is safe, 
 
 functional and always allowed. 
 
 A. 3-2 Information Storage and Retrieval Group 
 
 Responsibilities of this group include: 
 
 2.1 Display Consoles 
 
 2.2 Microform File 
 
 2.3 Switching Exchange between Consoles and Microform File 
 2.k Document Acquisition and Processing 
 
 2-5 OptJcal Disk Backup Store 
 2.6 Associated Programming 
 
 We speak here of two kinds of optical disk storage: the microform file 
 with photographic micro images and the backup store with digital encoding. 
 The trade-offs of which materials in an extensive file should be retained 
 in which form will doubtless remain an open question for some years. Accordingly 
 we have placed the responsibility for the engineering and use of both stores 
 under a common group. 
 
 In like manner the trade-offs between, console bandwidth and 
 image resolution at the microform file can be evaluated only by a 
 comprehensive group covering both display and file development. Also 
 questions of editing text by machine to be compatible with a lower resolution 
 image must be evaluated from actual console display experience. 
 
 Library science is commonly associated with this area. This 
 fieldcan suggest some workable standards and algorithms; these contributions 
 then fall either i) into document acquisition and processing, or ii) into 
 associated programming -- which here is interpreted to include fact retrieval. 
 
 A. 3-3 Computer Planning Group 
 
 This group has responsibility for: 
 
 3-1 Computer Central Processors 
 
 3-2 Fast Storage 
 
 3-3 Scratch Pad Backup Store 
 
 3-1+ Diagnostic Programming 
 
 • 5 System Programming 
 
 -76- 
 
There are obvious compatibility constraints between the Computer 
 Subsystem and the design of the File/Display Switching Exchange, the 
 Communications Net and the digitally encoded Microform File. In this 
 organization it is proposed that conventional system programming (assembler, 
 universal translator, executive system, etc.) be considered the natural 
 extension of computer engineering. 
 
 A. 3*^ Data Acquisition and Communications Group 
 
 For the development of a Information Resource Center, this 
 group has responsibility for 
 
 k.l Local Input/ Output Equipment 
 k.2 Communications Net 
 h.3 Associated Programming 
 
 Facsimile transmission is one of the easier ways to characterize and test a 
 communications net -- hence the association here with local input/ output 
 equipment . 
 
 If an attached automated laboratory is planned, then the 
 responsibilties of this group might be enlarged to include: 
 h.k Computer-type Macro-Modules 
 4-5 Instrumentation-type Macro-Modules 
 
 k .6 System Programming associated with the macro-module system/ 
 communications net. 
 
 Observe that a diffuse system of macro-modules can function as a piece of 
 a communications net. Accordingly the introduction of macro-modules extends 
 earlier concepts of a communications net. Also, in line with this 
 extension, data acquisition can include both preprocessing and process 
 control executed by the macro-modular net. 
 
 -77- 
 
ACKNOWLEDGMENTS 
 
 The list of individuals who have assisted in the formulation of 
 the views presented here is large. The contribution to the architectural 
 design by Ambrose M. Richardson, A. I. A., of Richardson, Severns, Scheeler 
 and Associates/Architects has been acknowledged by co-authorship. In addition, 
 however, I would like to acknowledge the stimulus he gave this entire 
 development by his encouragement, sympathetic belief and emphasis on design 
 over the three year period since I first discussed with him tentative plans 
 for an automated library. 
 
 Mr. R. C Amendola of the ILLIAC III project staff is co-author 
 of Chapter 5. In addition he contributed criticism to Chapter 1. I have 
 leaned upon his knowledge of industrial and optical design. 
 
 The present senior staff of the ILLIAC III Computer Project: 
 Professors Sylvian R. Ray, Manfred Paul and Dr. James L. Divilbiss contributed 
 much valuable criticism. In particular I would like to single out Professor 
 K. C Smith, University of Toronto, formerly a member of the ILLIAC III project 
 staff and now a consultant, for many discussions of the concept of an 
 Information Resource Center. I believe it was our common hope from the 
 earliest days of the ILLIAC III project that this machine could serve as a 
 prototype for the broader concept of an Information Resource Center. 
 
 Dr. Ben T. Williams, pathologist at Mercy Hospital (Urbana) has 
 guided much of my thinking about the medical information center application. 
 It is anticipated that the generalities of this report can be reduced to the 
 syntax, semantics and pragmatics of a later joint paper. The advise on 
 hospital administration and enthusiastic support of Father John Weishar, 
 Diocesan Director of Hospitals, Peoria, Illinois is also noted. 
 
 The concept developed here of an automated laboratory is in direct 
 response to a challenge from Professor Sol Spiegelman, Department of 
 Microbiology, University of Illinois. The potential of macro-modules was 
 discussed over the past 18 months with Professor Wesley A. Clark, Washington 
 University, St. Louis. 
 
 Professor James N. Snyder critically read the manuscript and 
 mmended many clarifications. Dr. John R. Pasta provided administrative 
 
 -78- 
 
assistance to shield me somewhat from the continuing administrative burden 
 of the ILLIAC III project while writing this report. 
 
 By listing the names above I do not wish to imply a concensus 
 of opinion on all arguments of the report. The style of the report is 
 evocative- -hopefully to stimulate serious deliberation. 
 
 -79- 
 

UNIVERSITY OF ILLINOIS URBANA 
 
 30112103707086