Object storage has traditionally been seen only in low level interfaces, visible only to kernels and filesystem code. However, if storage objects are made visible across a distributed system, it dramatically simplifies the construction of large storage systems that are flexible, expandable, and migrateable. The FOBS filesystem, consisting of a simple metadata layer containing pointers to storage objects on remote filesystems, demonstrates this concept. The layer of indirection allows for filesystems larger than any single disk, permits multiple filesystems to share common objects, and enables users to create and manage private namespaces. Experimentally, this work demonstrates that the overhead of indirection is low, a single client can write faster than disk speed, and multiple clients can harness aggregate disk throughput. FOBS scientific application is examined with case study of the filesystem as employed by a high energy physics experiment on a cluster of 32 nodes for over one year.