home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!gatech!darwin.sura.net!zaphod.mps.ohio-state.edu!pacific.mps.ohio-state.edu!linac!att!ucbvax!TRANSARC.COM!Craig_Everhart
- From: Craig_Everhart@TRANSARC.COM
- Newsgroups: comp.soft-sys.andrew
- Subject: Re: "slow networks"
- Message-ID: <kebZkqz0Bwwt4e=HEj@transarc.com>
- Date: 28 Aug 92 17:02:46 GMT
- References: <MebPm_m0ts4j4mfEtI@alw.nih.gov>
- Sender: daemon@ucbvax.BERKELEY.EDU
- Distribution: world
- Organization: The Internet
- Lines: 53
-
- Well, shucks. When Bill Cattey proposed jettisoning the message server,
- I assumed he was talking about jettisoning the funky message-server/MUA
- (mail user agent, e.g. cuis, vuis, messagess) split in favor of some
- newer protocol. Not a bad idea to explore, what with interesting
- mailbox-access protocols like IMAP and POP to check out. Certainly the
- MS/MUA split is cumbersome for linked-together MUA applications
- (messagesn, cuin, vuin--the normal case); what seemed appropriate for
- PC&Mac support in 1985 doesn't make that much sense today. No surprise
- here.
-
- I didn't realize that what Bill seemed to be targeting was the whole
- AMDS picture. (Yes, lots of folks worked on making that reliable,
- including guys like Mike Kazar making AFS work.) I just don't see how
- Bill's AMDS retrospective critique follows. Essentially, AMDS follows a
- client-server model, piggybacking on top of the AFS client-server model.
- AFS effectively provides a real good name service, allowing clients
- anywhere to find the correct server for the desired service.
-
- Given that Bill has never really lived with AMDS himself, I'm wondering
- how he comes to his conclusions. Perhaps the tenseness in the message
- server (the UI-support component) is the stuff that Bill finds
- objectionable. Indeed, it was largely a matter of taste, whether to put
- all that address validation support into the message server. I happen
- to like the support it offers. In trying to package AMDS+messageserver
- up for use at non-andrew.cmu.edu sites, the validation support becomes
- something of a nightmare for installers, but we tried to deal with this
- by making it reasonable for installers to turn all that stuff off. And
- it really can be helpful when it's in a working installation. The
- problem as I see it is that all the messageserver code to support
- address validation is nightmarishly complicated. Is that what Bill's
- complaint is really about?
-
- I guess I don't follow Bill's paragraph:
- Excerpts from internet.other.info-andrew: 27-Aug-92 Re: "slow networks"
- Bill Cattey@Athena.MIT.E (1456*)
-
- > The AMDS mail delivery method is fundamentally flawed. It is bad design
- > to require the delivery agent to have to deal with the possibility that
- > there are as many delivery agents as there are users all attempting to
- > write into a unix file system home directory to copy mail in and update
- > status simultaneously.
-
- Wish I did follow this statement, so I could form some opinion about it.
- Clearly, in any mail delivery scene, somebody serializes mail
- deliveries somewhere. What's the difference how or when that happens?
-
- I wouldn't claim that AMDS is flawless, or that there aren't parts that
- should be re-worked or withdrawn. (Or that the messageserver
- name-validation stuff is simple.) I just don't understand this
- criticism--whether it's based on some new religion whose evangels
- haven't reached this shore, or what.
-
- Craig
-