Just as it is possible to mount more than one file system on a syntactic mount point [rp:93], it is also possible to mount more than one name space on a semantic mount point. HAC treats each such name space as an independent entity. The scope of queries asked within a multiple semantic mount point is simply a union of the scope provided by each mounted name space. Queries are evaluated independently in each name space and their results are treated as disjoint sets of symbolic links. The only restriction is that all name spaces mounted on a multiple semantic mount point must be accessible via the same query language. (Currently, HAC does not deal with overlapping name spaces or data, i.e., it does not resolve cases where two symbolic links might actually point to the same remote file, or to similar files.)
For example, suppose that we want to cover Remote and Local at the same time. All we need to do is create another semantic mount point s.Local as a multiple semantic mount point! There is no problem of cyclic reference here, because s.Local is just an interface to a CBA mechanism; it does not provide CBA on its own.
Syntactic and semantic mount points can be combined in various ways to share information by both name and content. Getting back to our fingerprint example, we may have access to a digital library with scientific articles. We can add a semantic mount point associated with a query for "fingerprint" (or a more complex query), thus ensuring that our knowledge of the subject is up to date (at least with the library). There will probably be other sources as well, and we may need to form different queries depending on the source. We may want to have syntactic mount points to all these sources and search there manually once in a while, but in addition HAC allows a user to build remote semantic directories for each source (or one for all of them), and have a better access to and better integration with this information. For example, one can "remove" certain results of no interest, add comments, add results from other places, etc. Other users (e.g., coworkers on the same project) can use syntactic mount points to browse through one user's personal classification (instead of doing the searches themselves) and retrieve relevant information. It is also possible to collect the names, queries and query-results of many semantic directories of many users in a central database that itself can be indexed and searched. Users can browse and search this database and find others who have similar tastes as they have. This may help them find what they are looking for even more quickly. Finally, users can add their favorite books, articles, memoirs, short-cuts to other information, etc., to their personal HAC file systems, index and search them, and export their file systems as mini-digital libraries to others. To conclude: semantic mount points give us a powerful new way to access the semantic aspect of information. They can be combined with syntactic mount points to yield a rich set of primitives for sharing information in a distributed system.