6.826—Principles of Computer Systems 2004 Handout 13. Paper: Semantic File Systems 1 13. Paper: Semantic File Systems The attached paper by David Gifford, Pierre Jouvelot, Mark Sheldon, and James O’Toole was pre-sented at the 13th ACM Symposium on Operating Systems Principles, 1991, and appeared in its proceedings, ACM Operating Systems Review, Oct. 1991, pp 16-25. Read it as an adjunct to the lecture on naming6.826—Principles of Computer Systems 2004 Handout 13. Paper: Semantic File Systems 2 Semantic File Systems David K. Gifford, Pierre Jouvelot1, Mark A. Sheldon, James W. O’Toole, Jr. Programming Systems Research Group MIT Laboratory for Computer Science Abstract1 A semantic file system is an information storage system that provides flexible associative access to the system’s contents by automatically extracting attributes from files with file type specific transducers. Associative access is provided by a conservative extension to existing tree-structured file system protocols, and by protocols that are designed specifically for content based access. Compatibil-ity with existing file system protocols is provided by intro-ducing the concept of a virtual directory. Virtual directory names are interpreted as queries, and thus provide flexible associative access to files and directories in a manner com-patible with existing software. Rapid attribute-based access to file system contents is implemented by automatic extrac-tion and indexing of key properties of file system objects. The automatic indexing of files and directories is called “semantic” because user programmable transducers use information about the semantics of updated file system ob-jects to extract the properties for indexing. Experimental results from a semantic file system implementation support the thesis that semantic file systems present a more effec-tive storage abstraction than do traditional tree structured file systems for information sharing and command level programming. 1 Introduction We would like to develop an approach for information storage that both permits users to share information more effectively, and provides reductions in programming effort and program complexity. To be effective this new approach must be used, and thus an approach that provides a transi-tion path from existing file systems is desirable. In this paper we explore the thesis that semantic file systems present a more effective storage abstraction than do traditional tree structured file systems for information shar-ing and command level programming. A semantic file sys-tem is an information storage system that provides flexible This research was funded by the Defense Advanced Research Pro-jects Agency of the U.S. Department of Defense and was monitored by the Office of Naval Research under grant number N00014-89-J-1988. 1 Also with CRI. Ecole des Mines de Paris. France. associative access to the system’s contents by automatically extracting attributes from files with file type specific trans-ducers. Associative access is provided by a conservative extension to existing tree-structured file system protocols, and by protocols that are designed specifically for content based access. Automatic indexing is performed when files or directories are created or updated. The automatic indexing of files and directories is called “semantic” because user programmable transducers use information about the semantics of updated file system ob-jects to extract the properties for indexing. Through the use of specialized transducers, a semantic file system “under-stands” the documents, programs, object code, mail, im-ages, name service databases, bibliographies, and other files contained by the system. For example, the transducer for a C program could extract the names of the procedures that the program exports or imports, procedure types, and the files included by the program. A semantic file system can be extended easily by users through the addition of special-ized transducers. Associative access is designed to make it easier for us-ers to share information by helping them discover and lo-cate programs, documents, and other relevant objects. For example, files can be located based upon transducer gener-ated attributes such as author, exported or imported proce-dures, words contained, type, and title. A semantic file system provides both a user interface and an application programming interface to its associative access facilities. User interfaces based upon browsers [Inf90, Ver90] have proven to be effective for query based access to information, and we expect browsers to be offered by most semantic file system implementations. Application programming interfaces that permit remote access include specialized protocols for information retrieval [NIS91], and remote procedure call based interfaces [GCS87]. It is also possible to export the facilities of a semantic file system without introducing any new interfaces. This can be accomplished by extending the naming semantics of files and directories to support associative access. A benefit of this approach is that all existing applications, including user interfaces, immediately inherit the benefits of associa-tive access. A semantic file system integrates associative access into a tree structured file system through the concept of a6.826—Principles of Computer Systems 2004 Handout 13. Paper: Semantic File Systems 3 virtual directory. Virtual directory names are interpreted as queries and thus provide flexible associative access to files and directories in a manner compatible with existing soft-ware. For example, in the following session with a semantic file system we first locate within a library all of the files that export the procedure lookup_f ault, and then fur-ther restrict this set of files to those that have the extension c: % cd /sfs/exports:/lookup_fault % Is -F virtdir_query.c@ virtdir_query.o@ % cd ext:/c % Is -F virtdir_query. c@ % Semantic file systems can provide associative access to a group of file servers in a distributed system. This distrib-uted search capability provides a simplified mechanism for locating information in large nationwide file systems. Semantic file systems should be of use to both indi-viduals and groups. Individuals can use the query facility of a semantic file system to locate files and to provide alterna-tive views of data. Groups of users should find semantic file