preece@urbana.mcd.mot.com (Scott E. Preece) (02/10/90)
I am currently looking at ways to support further typing of UNIX system
objects (files, processes, etc.). I would appreciate pointers to
previous work in the area and to existing systems that support richer
models of system objects (particularly those that support attaching
arbitrary collections of named attributes to objects, as I understand
the Macintosh Resource fork allows).
I'm also interested in comments on how such features might be or have
been used, what a minimal ("in the spirit of UNIX") approach might look
like, and what you think might be critical aspects of such features (if,
for instance, your experience is that supporting a LIFO list of
properties works much better than a flat set, let me know).
Mail responses would be great and I promise to summarize whatever I get;
don't post unless you really think everyone would be interested.
Thanks,
scott
--
scott preece
motorola/mcd urbana design center 1101 e. university, urbana, il 61801
uucp: uunet!uiucuxc!mcdurb!preece, arpa: preece@urbana.mcd.mot.commogul@decwrl.dec.com (Jeffrey Mogul) (02/13/90)
In article <1198@darkstar.ucsc.edu> preece@urbana.mcd.mot.com (Scott E. Preece) writes: > >I am currently looking at ways to support further typing of UNIX system >objects (files, processes, etc.). I would appreciate pointers to >previous work in the area and to existing systems that support richer >models of system objects (particularly those that support attaching >arbitrary collections of named attributes to objects, as I understand >the Macintosh Resource fork allows). If I might forgiven some self-promotion, you should read my dissertation "Representing Information About Files", available from the Stanford University Computer Science Department as Report No. STAN-CS-86-1103. They will charge you for this; I don't have many copies left so I'd rather you got one from Stanford. Note that I do not really believe that the idea of attaching arbitrary information to files would work well in the traditional Unix environments. It might be reasonable in a system that provided a Unix-like interface along with a richer one. Also note that I've basically ignored this whole area since finishing my thesis 4 years ago. -Jeff
brian@june.cs.washington.edu (Brian Bershad) (02/15/90)
Since everybody is busy forgiving Jeff's self-promotion, perhaps they could forgive this as well. For a general way to bind property lists to files, you might check out a paper from the Spring 88 Computing Systems called "Watchdogs -- Extending the UNIX File System" (it also showed in the Winter Usenix of that year).