jgp@moscom.UUCP (Jim Prescott) (07/30/87)
It occurs to me that OtherRealms is not the only place where it is desirable to try and insure that an old article stays around until a new one shows up. Things like arbitron results, newuser documents and uucp maps all have a lifespan that ends when the next posting of the info is received. What makes OtherRealms look different is that the length of time between postings necessitates very long expiration times. With OtherRealms the idea is to keep at least one version around, with groups like newusers the idea is to keep at most one version (the most current). How about a new header line of the form: Supersedes: message_id [message_id ...] When an article comes in with a Supersedes line news posts it normally and then does a cancel on the message_ids indicated. As an example, a weekly posting could be given a one month expiration date but would carry a Supersedes line for the previous 4 weeks messages. Even if a site misses one or two postings they will always have only the most recently received copy available. This is similar to just having the people who make the periodic postings send a normal cancel for the old copies but has the advantage that the cancel and the replacement article always arrive together. It is also more efficient since in the above example the poster would need to send 4 cancel messages every week to get the same protection against articles being lost. Some problems I can see with this are: - It is more bookkeeping to track the old message_ids. Since most of the periodic postings are probably largely automated this shouldn't be too bad. - It might lead to longer expiration times being used with the intent that the article would be superseded before expiring anyway. This will annoy the people whose software doesn't understand the new header line. Some benefits are: - Savings of disk space since old messages will be removed as soon as they become obsolete instead of waiting for expire. - One of the groups this would help would be newusers, anything that cuts down the volume there without losing info will make it more likely that people will consider reading the stuff in the first place. - It becomes easier to make sure that even sites with flaky newsfeeds always have a recent copy of your article. Any comments? -- Jim Prescott seismo!rochester!moscom!jgp or moscom!jgp@cs.rochester.edu
roy@phri.UUCP (Roy Smith) (07/31/87)
In article <1026@moscom.UUCP> jgp@moscom.UUCP (Jim Prescott) writes: > How about a new header line of the form: > Supersedes: message_id [message_id ...] First off, yes, I think that's a wonderful idea. Actually, you can get the same effect (I don't believe I'm actually suggesting this) by putting a very long expriation date on the original article and sending out the replacement and a cancel control message for the first one at the same time. It's not as neat, but has the (non-trivial) advantage that it doesn't require any changes to the news code. Second, I think I owe the net in general, and Chuq in particular, and apology. Over the past few days, I posted rather a number of articles on the subject of the OtherRealms Expires: headers. While I still stand by everything that I said, the manner in which I said it was inappropriatly harsh and antagonistic. -- Roy Smith, {allegra,cmcl2,philabs}!phri!roy System Administrator, Public Health Research Institute 455 First Avenue, New York, NY 10016
matt@oddjob.UChicago.EDU (Matt Crawford) (08/02/87)
In article <1026@moscom.UUCP> jgp@moscom.UUCP (Jim Prescott) writes:
) How about a new header line of the form:
) Supersedes: message_id [message_id ...]
) When an article comes in with a Supersedes line news posts it normally
) and then does a cancel on the message_ids indicated. ...
) Any comments?
I think this is a grade-A #1 good idea. It is general - I can
already think of lots of situations in which it would be useful.
The network map files, Baby Doc Werner's AIDS postings,
positions in PBM games, newsgroup charters posted to the groups,
answers to common question in newsgroup X, and plenty more.
One caveat: it should be no easier to forge a "Supersedes" field
than a "Cancel" message. Probably this means only articles from
the same poster can be superseded.
________________________________________________________
Matt University matt@oddjob.uchicago.edu
Crawford of Chicago {astrovax,ihnp4}!oddjob!matt
biep@cs.vu.nl (J. A. "Biep" Durieux) (08/03/87)
In article <1026@moscom.UUCP> jgp@moscom.UUCP (Jim Prescott) writes:
"(...)
"How about a new header line of the form:
" Supersedes: message_id [message_id ...]
"When an article comes in with a Supersedes line news posts it normally
"and then does a cancel on the message_ids indicated. (...)
"Some problems I can see with this are:
" - It is more bookkeeping to track the old message_ids.
" - It might lead to longer expiration times being used (...).
"Some benefits are:
" - Savings of disk space (...).
" - One of the groups this would help would be newusers (...).
" - It becomes easier to make sure that even sites with flaky
" newsfeeds always have a recent copy of your article."
Sounds good! One remark: only the poster of the original article should
be able to post a superseding one.
About the bookkeeping: rn should have an option (no DeLorean this time,
Larry :-)) to do this easily (like the cancel command now). It would be
nice to be able to supersede instead of just cancel an article one wrote
and one isn't really happy with, or an article one isn't sure about it
really went out. Perhaps the superseding article should be cross-posted
to control.
--
Biep. (biep@cs.vu.nl via mcvax)
Some mazes (especially small ones) have no solutions.
-- man 6 maze
cds@root.co.uk (Chris Seabrook) (08/07/87)
Another advantage of allowing a poster to superceded an old article would be in the sources groups. How often do you see a 're-posting part 19/57 - previous version truncated' type message ? With this system the re-posting could supercede the malformed previous version.
mangler@cit-vax.Caltech.Edu (System Mangler) (08/10/87)
In article <1026@moscom.UUCP> jgp@moscom.UUCP (Jim Prescott) writes: >How about a new header line of the form: > Supersedes: message_id [message_id ...] >When an article comes in with a Supersedes line news posts it normally >and then does a cancel on the message_ids indicated. (...) But does the replacement article get the old article number, or a new number? Which you do depends on whether you think readers of the old message should see the new article, or not. It sure would be nice to go back and correct errors in a posting without anyone having to see the article a second time. Of course if you *want* them to see it again, you can cancel and start over... Perhaps the header should be "Replaces: message-id" and it should overwrite the old article without removing anything? Don Speck speck@vlsi.caltech.edu {ll-xn,rutgers,amdahl}!cit-vax!speck
rees@apollo.uucp (Jim Rees) (08/12/87)
How about a new header line of the form: Supersedes: message_id [message_id ...] When an article comes in with a Supersedes line news posts it normally and then does a cancel on the message_ids indicated. I liked this idea so much, and it's so easy to do, that I went ahead and implemented it. It seems to work great. I have posted the source to news.software.b.