Path: utzoo!utgpu!watmath!iuvax!ux1.cso.uiuc.edu!uwm.edu!rpi!tale
From: tale@pawl.rpi.edu (David C Lawrence)
Newsgroups: news.admin
Subject: Re: News Versions, another approach
Message-ID: <1989Nov16.234445.16469@rpi.edu>
Date: 16 Nov 89 23:44:45 GMT
References: <1989Nov15.213552.9499@mks.com>
Organization: Rensselaer Polytechnic Institute, Troy NY
Lines: 30

In <1989Nov15.213552.9499@mks.com> egisin@mks.com (Eric Gisin) writes:
Eric> This analysis differs from the version-control-message method in
Eric> two ways.  First, it only recognizes sites that have posted in
Eric> the last two weeks.  Second, it also recognizes sites that do
Eric> not respond to version messages, which makes this method more
Eric> reliable. Many C news sites don't seem to respond.  It is also
Eric> including many mailing list gateways.

It is a little more reliable as far as finding those sites goes, as
long as someone has posted a message from there.  It is a little
inaccurate though because some sites change the format of their
Message-IDs in spite of whatever news version they are running.  (Is
CMU running B or C News?  They've got some hairy M-IDs.)  As an
example, if you had done this in late late September then Rensselaer
would have been reported as both a C News site and a B News site even
in spite of the fact that we were running C News the whole time; in
fact, in the very near future we will be consistently mis-represented
as a B News site by this method.

Then there is all of the other M-IDs which find their way into news,
from either mailers (as you mention) or from client programmes like
GNUS.  Add to that all of the extra site names which appear in M-IDs
(is rpi.edu one site or is it six for each *.rpi.edu sub-domain?  Or
is it six hundred for each *.*.rpi.edu machine?) and the heuristic
begins rapidly failing.  It seems like a good general idea but I think
it needs refinement.

Dave
-- 
 (setq mail '("tale@pawl.rpi.edu" "tale@ai.mit.edu" "tale@rpitsmts.bitnet"))