ietf-nntp Notes from the IETF41 meeting

[Dave Barr]

In message <199804031746.LAA04857 at amusingly.bogus.iastate.edu>, John Hascall wr
ites:
>      IHAVE xxx   ------------->
>                 <-------------   "DONT WANT THAT"
>      WHYNOT      ------------->
>                 <-------------   REASON(S)

This is an interesting idea.

How long do "REASON"'s apply?  This would be an ever-increasing list
of reject reasons, which the sending site would have to manage.
Is there a way for a site to say "oh, by the way, the newsgroup X I was
rejecting before, I can accept that group now".  If they expire
after a day or so, then we lose much of the gains.  If they take too long
to expire, new groups would take longer to begin propagating.

My concern is that a large number of articles are thrown away because
of typos in Newsgroups: and Distributions: lines.  Over months/years,
the "don't send this newsgroup/distribution to site x" list is going to
be quite large if not trimmed.

I still like the idea of a scheduled system by which the sending
site does a "LIST FEEDENTRY", and the site spits back a structured
list of what the newsfeeds entry should be.  Sites can opt to have
this be "here's my active file, send me only these groups, within these
distributions I accept", or they can opt for a more structured newsfeeds-
style system where they just say "here's the heirarchies and distributions
I carry, send me anything that matches", or they could do a combination
("here's an explicit list of alt groups I want, plus these heirarchies,
etc").

I realize this is a non-trivial system which needs to be written,
but it's one which can be easily tacked onto most systems without
source code changes by doing something out-of-band.  Making a real
NNTP command for all this would of course require source changes.

My other concern is NNTP is already a very lock-step protocol.
I realize there will probably be a streaming counterpart to this,
but I wish that the news transport datastream would try to conserve
all its bandwidth towards moving bits across the wire.  While this
would conserve bandwidth in the end, I think the other system would
conserve more.  The FEEDENTRY stuff would be done off-peak when bandwith
is cheap, and leave the peak times for maximum news flow.

--Dave