Minutes of NNTP-Managers BOF at Usenix

[Rich Salz]

Here's the overdue minutes of the NNTP-Managers BOF at Usenix.  I don't
think I missed anything, if so let me know.

A lot more showed up than I first expected -- around 40 or 50, rather than
20 or 30.  A possible drawback to this was that there were too many people
for us to go around and introduce ourselves; a definite positive result
was that I didn't have to use too many of my prepared jokes.

At the beginning of the meeting I passed out copies of a map that I made
based on the "Distribution: inet" sendsys I did.  If you want a copy, you
can pick it up from uunet.uu.net:~/uumap/m.inet-sys.ps.Z.  If someone is
really interested in the programs and raw data, let me know.  I also have
hardcopy maps of many of the regionals.  This will involve photocopying
and postage, so you better be VERY interested (or have a good bribe :-) if
you want those.  The most interesting fact from my map is that there are
a lot of cross-country links!

As expected, most of the talk was non-directed discussion about whether or
not we should organize the NNTP feeds, and whether we need or want more
structure at all.  This question can be broken down into two levels, the
NSFNet backbone level and the regional network level.  The regionals are
clearly on their own, and the only interest or concern is news flow over
the NSFNet, and similar government-sponsored (your tax dollars at work)
networks.

The discussion was mostly anecdotal.  There was some talk about optimizing
the NNTP protocol for fewer TCP packets.  A couple of folks seemed to have
memorized PING round-trip times (the modern-day equivalent of knowing the
statistics on the back of baseball cards, perhaps).  On the political
front, there are two major arguments:  we should use the network as much
as possible, "flooding" it with redundant connectivity in order to show
everyone that the networks are being used; or, we should try to keep a
relatively low profile so that "they" don't come and kick us off.  Various
guesses and conjectures were made as to the amount of traffic NNTP
actually takes up.  One set of measurements (claimed to have come from
Peter Honeyman) said it was #5, after SMTP, Telnet, Name Service, and
FTP.  Someone later told me that the Steve Wolff (head of NSFNet) has NNTP
at 15% of the traffic along the NSFNet backbone, and they have no problem
with that.  (I'm not naming the someone because I didn't ask their
permission to quote them directly in this public forum.)

In short, everyone pretty much agreed that the current anarchistic
hodge-podge is fine technically and politically.  At some point this may
change, and we might have to become organized.

The next topic was running NNTPXMIT.  Most people run it every 30 to 60
minutes, a few run it every one to 10 minutes, and Scott Bradner at
Harvard runs it continuously.  **Care to post your script that does this,
Scott?**  After pleading from Erik Fair, most people agreed to run it more
often.  **Please run NNTPXMIT at least every 10 minutes.**  A couple of people
are working on newer and faster variants, which are in various stages of
testing.

The next topic was transferring "biz" and other commercially-oriented
services.  Before the meeting I spoke to Jane Caviness at NSF.  She works
for Steve Wolff and is the Program Director of NSFNet.   She said that the
NSF is working on a usage policy, and that it's been in the hands of the
lawyers for some time, and is expected any week now.  I later heard that
it's been "any week" for several months now.  The policy will basically
state that most advertising is not appropriate, and that commercial use is
only valid when it is supporting government-funded R&D.  For example,
unsolicited "sales ads" are not allowed, but an email reply from a sales
office to a general RFP is fine.

The regional nets make their own rules, of course, but are responsible for
making sure that nothing inappropriate leaks out to the NSFNet backbone.
Consequently, and because it's in general easier, most regionals have said
that they will be adopting the NSFNet policies.  The "usage policy"
questions have also been debated a great deal among FARNet, the federal
network organization.  [If what William Wulf said during his keynote is
correct, than all of the national networks will be merged into one
national R&D network in the next five to 10 years.  This would probably
mean we'll all be operating under NSFNet rules.]

As I said at the meeting, my impression is that the rules are gonna get
more strict.  **Regardless, it is probably not a good idea to carry
"biz.all" over the NSFNet backbone.**  There was lots of dicussion about
Clarinet and the In Moderation Network.  Templeton has a proposal in
before Wolff (not Wulf:-) that would formally okay the transfer of his
stuff over NSFNet.  Until then, it seems to be that as long as a site
requests such a feed, it's for their work, so it is probably okay.
The statement that as long as it is in direct support of an organization's
research work it's okay, might have interesting implications for the
UUCP maps

Nothing is certain until NSF policy is published.

There wasn't much interest and discussion in NNTP on BITNET, VMS, or other
strange creatures.  A couple of people wanted to know NNTP flow with
UUNET.  Several people feed it, but only one or two feeds come out.

Oh yeah:  I also put out a list for people to get added to the nntp-managers
mailing list.  I sent the names along to the maintainer.

I liked the BOF, thought it went pretty well, and got similar feedback.
I'll probably offer to run one next June, too.
	/r$
-- 
Please send comp.sources.unix-related mail to rsalz at uunet.uu.net.
Use a domain-based address or give alternate paths, or you may lose out.