davecb@yunexus.UUCP (David Collier-Brown) (01/02/90)
>In article <89Dec31.171430est.2251@neat.cs.toronto.edu> moraes@cs.toronto.edu (Mark Moraes) writes: >> For us, centralizing .newsrcs is technically hard (we prefer less >> interdependency between our servers, not more) and politically >> impossible. >[ And other stuff about how it isn't such a hot idea at his site.] tale@cs.rpi.edu (David C Lawrence) writes: >Indeed. This site is quite the same way; nevertheless the model is >acceptable for many other sites on the net. Excuse me, but you don't really have to centralize .newsrcs! You need only keep them "with" the machine that the particular news spool is on. [True centralized .newsrcs would be a catastrophy at here at Orc, as they would introduce a single point of failure for the whole news system: we have quite enough single-point failures, thank you!] The "Forum" system kept a construct much like a directory of .newsrc files (it was a segment of records), and worked quite well because the format was well-known and could be manipulated by a standard set of utility routines: the reader was quite independant of the format. It tended to trivialize the expirey problem, since all the information about who read group X was localized "near" group X. What was required was either 1) security to keep one user from munging another's file, or 2) reliability in the newsrc-manipulating code. [or both] In an NFS environment with NNTP posting, the latter can be faked by having each person have a rw mount of the directory containing the .newsrc files, having the person's real .newsrc be a symbolic link (or contain a reference line for non-symlink NFS systems if such exist) and have the add-user-to-system script generate a long unique name for the newsrc file, perhaps their full name... This is clearly a cute hack (:-{), but it does make the newsrc files acessable to a usage-collector... --dave -- David Collier-Brown, | davecb@yunexus, ...!yunexus!davecb or 72 Abitibi Ave., | {toronto area...}lethe!dave Willowdale, Ontario, | Joyce C-B: CANADA. 416-223-8968 | He's so smart he's dumb.
cgf@ednor.bbc.com (Chris Faylor) (01/03/90)
Instead of reinventing a new ".newsrc" format, why not just develop individual filters for the formats used by the finite number of newsreaders that are out there? The output for the filter is the input for the new-improved expire. An individual user could 'register' the newsreader he was using so that the expire process would know what filter to apply to his ".newsrc". Any future newsreader writers could also be expected to provide filters for their programs. Or is this too slow? Seems like we are already talking s..l..o..w.. if we are talking about reading each user's file and generating tables from them. -- Chris Faylor cgf@ednor.bbc.com