We have shown in this paper that it is possible, with reasonable overhead, to combine name-based and content-based access to files at the same time, while preserving the main benefits of both methods. We identified several scope and consistency problems, suggested solutions, and described an implementation. This is obviously not the last word on this topic. More work needs to be done to convince people to add a major paradigm to their daily arsenal of tools. A major missing piece is usability testing, which we have not performed to date. Only when a working version is widely distributed and used can we determine how beneficial is this approach.
The implementation described in the paper was geared towards personal file systems, and as is it is not scalable to very large file systems (e.g., Internet wide). This is true for many existing file systems. In particular, our decision not to modify anything at the kernel level (so the system will be easier to distribute) adds quite a bit to the overhead. Implementing our ideas in the context of a large scale Intranet, for example, will be a major next step. We believe it is possible and very desirable. The current situation of server-based search facilities and user-based file systems with almost no connection between them can be and should be improved.