🚀 go-pugleaf

eastick@me.utoronto.ca (Doug Eastick) writes:

>mailrus first, then to jarvis.  Could watmath be using NNTP over Onet
>to iuvax?  If that's the case, why not talk to jarvis instead?

Assuming Waterloo runs NNTP this would make perfect sense, but it can't happen
until you guys finally take an hour and install NNTP (ditto for utcs) so
jarvis can do less batching (radio.astro could take over the feed, but it
sounds silly to cross n campus gateways when utme is one hubnet hop away from
CSRI).  If there were more NNTP sites on campus, it would be much easier to
propate articles faster to York, Waterloo and other ONET sites, as it would be
possible to share the load without paying the price of important added latency.

If indeed Waterloo is fed from Illinois over the ONET link, then essentially
all the data is coming over the US link twice.  Even if there was ample
bandwidth (which is not the case, given that the current link is 38.4 Kbaud),
it seems silly not to have those two sites talk over NNTP, which would at
least prevent some of the double transmissions.  The optimal scenario is to
have one "receiving hub" for each independant connection from ONET to other
networks.  Should the link to RISQ in Que'bec (and hence alternate access to
the US via NYSERNET) materialize, another ONET site could be fed from there,
providing redundancy without competing for the bandwidth on the outside-world
links.  Given the ONET topology,  it would be conceivable to have jarvis.csri
keep its NNTP feeds over the NSFNet connection, with ugw.utcs or gpu.utcs
connecting to a site over on RISQ.  Then news could be efficiently pushed to
the rest of ONET.

I think it would be time for the news admins on sites with access to ONET to
start discussing ways of rationalize our feeds (NNTP has the potential of
changing certain things in important ways).

Jean-Francois Lamy               lamy@ai.utoronto.ca, uunet!ai.utoronto.ca!lamy
AI Group, Department of Computer Science, University of Toronto, Canada M5S 1A4