rsalz@bbn.com (Rich Salz) (04/01/88)
Bob Page:
=Anyway, the backbone can up and die for all the net cares. It just
=means people will have to make decisions (or not) for themselves
Greg Woods:
= Yeah, AND they have to find some new sites willing to foot the phone
=bills/CPU cycles/disk space to transmit everything. The result of that
=is that there is still a backbone, it's just a different group of sites.
I checked out Gene Spafford's latest "What's the Backbone" article. Greg
metions the three three primary criteria for backboneage: phone bills,
CPU, and disk space. Just how relevant are these things, in these days of
easy-access networks (such as NSFNET) and NNTP? I'm not sure, but I tried
an interesting experiment: I tried to connect to the NNTP port on as many
of the US machines on Gene's backbone map as I could. Here's what I got:
ARPA/Internet sites I could successfully connect to for NNTP
ames
cmcl2
decvax
gatech
hao
hplabs
husc6
mcnc
mit-eddie
rutgers
ucbvax
ucsd
uunet
uw-beaver
ARPA/Internet sites where a connection to port 119 was refused:
bellcore
decwrl
ukma (map shows news uunet link)
Sites that are apparently only UUCP:
amdahl (map shows news uunet link)
clyde
ihnp4
linus
nbires (map shows uunet mail link)
philabs (map shows uunet mail link)
tektronix
ulysses
Sites known to be dead as of right now:
cbosgd (presumably cb-att is filling this spot right now, but
lots of folks fled when cbosgd died...)
cae780
There are definitely some things wrong with my simplistic analysis. For
example, I didn't post a sendsys to the sites involved (I suppose I
could, via NNTP, but I'm about to leave for the day :-), so I don't know
how much time, money, and cycles are spent in the old-fashioned
UUCP/phonecall/compress method of transmitting netnews. I'm also
completely wrong ignoring conventional UUCP email traffic.
Still and all, the above list make me draw some conclusions that I really
and honestly don't like. In increasing order of my dislike, they are:
The growth of the Internet, in particular the massive explosion of the
NSFNET, and the astounding success of UUNET have drastically changed
the structure of Usenet.
The avowed concern of the backbone for "time, money, cycles, and
space" is -- for the majority of sites -- not really an issue for
them; the continual restatement now feels to me like paternalism.
Without some re-evaluation of its nature, goals, and priorities
the "backbone" is doomed to become as relevant to Usenet, as
the appendix is to digestion.
/r$
--
Please send comp.sources.unix-related mail to rsalz@uunet.uu.net.david@ms.uky.edu (David Herron -- One of the vertebrae) (04/02/88)
In article <571@fig.bbn.com> rsalz@bbn.com (Rich Salz) writes: >I checked out Gene Spafford's latest "What's the Backbone" article. Greg >metions the three three primary criteria for backboneage: phone bills, >CPU, and disk space. Just how relevant are these things, in these days of >easy-access networks (such as NSFNET) and NNTP? I'm not sure, but I tried >an interesting experiment: I tried to connect to the NNTP port on as many >of the US machines on Gene's backbone map as I could. Here's what I got: ... >ARPA/Internet sites where a connection to port 119 was refused: ... > ukma (map shows news uunet link) A small piece of information for you. You probably tried to connect to "a.ms.uky.edu" because of the alias "ukma = a.ms.uky.edu" in our map entry. What you don't know is that our NNTP server is run from an entirely seperate machine -- e.ms.uky.edu. Our news is physically stored on "e" with "a" being used as the interface to BITNET, and "g" being used as the interface to UUCP. (except the uucp software on "g" thinks that its name is "ukma" ... software is such fun :-) ... ). > The growth of the Internet, in particular the massive explosion of the > NSFNET, and the astounding success of UUNET have drastically changed > the structure of Usenet. I know that we would never have been able to become part of the backbone if it weren't for NSFNET. Before NSFNET we were limping along on a pair of news feeds over BITNET (one which was from an IBM machine and all the articles we receive from there get horribly munged), plus one or two partial feeds with UUCP. As much as our phone budget could afford. It was a bad time. I agree with you. NSFnet has created at least one backbone site. > The avowed concern of the backbone for "time, money, cycles, and > space" is -- for the majority of sites -- not really an issue for > them; the continual restatement now feels to me like paternalism. My time is important. The more time I spend administering news the less time I have to do other things which are in reality very important. My time does cost the school some money. "e" gets slow whenever we get 3-4 NNTP connections running at a time. we've got a 100+ meg partition for news. nothing to sneeze at. I'd say that all those things are important to this place. Or were you implying that those things aren't important for most sites, but are for sites that have lots of feeds? That's probably true. One of the important duties for sites that have lots of feeds is to streamline and the software. It *is* important here that the software be more efficient (the constant exec'ing of rnews from nntpd is a killer) at this site and surely at others. > Without some re-evaluation of its nature, goals, and priorities > the "backbone" is doomed to become as relevant to Usenet, as > the appendix is to digestion. Occasional re-evaluation is good for the soul. > /r$ -- <---- David Herron -- The E-Mail guy <david@ms.uky.edu> <---- or: {rutgers,uunet,cbosgd}!ukma!david, david@UKMA.BITNET <---- <---- I don't have a Blue bone in my body!
brad@looking.UUCP (Brad Templeton) (04/02/88)
In fact, in an odd way it's Telebit Technologies that will probably deal the end to the backbone concept, in combination with uunet type sites. With a Telebit you can send the 70 megs of monthly news, compressed to about 35 megs in about 7 hours. I know that AT&T's reach out America is $9/hour, that's $63/month, twice that for a double long distance feed. That's to anywhere in the USA, once you fork out the $700 a backbone site pays for one of those modems. With Telenet PC-persuit, if you have it, it's $25 for the whole month for everything. Of course, backbone sites today carry a very large mail burden, but that would change with wider distribution of long distance duties. To me, it's amazing. I remember how we used to talk about $100K phonebills for decvax and how it would all come crumbling down for financial reasons. It's astounding what technology has done. Now the big cost in news is not the physical cost, but the cost in time spent reading it. [ Note that the above analysis is simplistic, not counting PC-persuit. It assumes transmission only at night, and perfect batching for 100% uucp efficiency. If we cared, we could get that 100% efficiency, but it just doesn't matter. Even at only 50% efficiency it's no longer a major bother if you use strictly telebits, which everybody isn't using yet. ] [ This is not to belittle in any way the incredible contribution of dollars that some backbone sites have made. I still wonder how they did it. ] -- Brad Templeton, Looking Glass Software Ltd. - Waterloo, Ontario 519/884-7473
mark@cblpf.ATT.COM (Mark Horton) (04/09/88)
In article <571@fig.bbn.com> rsalz@bbn.com (Rich Salz) writes: >Sites that are apparently only UUCP: > ihnp4 >Sites known to be dead as of right now: > cbosgd (presumably cb-att is filling this spot right now, but > lots of folks fled when cbosgd died...) A small correction, please: Both ihnp4 and cbosgd are being phased out, although Tim's message did just mention cbosgd. They are being replaced with att-ih and att-cb. Netnews is already flowing over the new machines, although most email is still going via ihnp4 and cbosgd. ihnp4 and cbosgd are both still up and running OK. Rather than publicize names like att-cb and att-ih, we're going to call them both "att" and arrange that they appear to be one machine. Work is progressing right now to accomplish this. This is why neighbors of att-cb are being contacted to instead become neighbors of att. Neighbors of att-ih should expect to be called soon, and may wish to add "att" to their Systems files in addition to att-ih. The real message to most of you is this: cbosgd and ihnp4 are still up, but will be phased out over the next few months. The plug could be pulled around July 1. A new machine called "att" is coming up now. The UUCP map will be updated to advertise "att" within a week or two. Those of you with names like "cbosgd" or "ihnp4" hardwired into your routing tables, or your heads, should think about "att" instead. Those of you with such names in your .signature file or in your product literature should find alternate routes (we recommend attmail at 1-800-MAIL672, but of course there are many alternatives) Eventually (early 1989) we anticipate other many other AT&T gateway machines (such as clyde, ulysses, mtune, etc) being folded into the "att" structure, so that we present a uniform appearance to the rest of the world. Local dialups in each of the areas are anticipated; you won't have to call Chicago from New Jersey. Mark Horton wearing my AT&T hat