Xref: utzoo news.misc:1485 news.config:628
Path: utzoo!utcsri!jarvis.csri.toronto.edu!utgpu!water!watmath!clyde!bellcore!faline!thumper!ulysses!ucbvax!decwrl!purdue!spaf
From: spaf@cs.purdue.EDU (Gene Spafford)
Newsgroups: news.misc,news.config
Subject: Re: The USENET Backbone (Updated: 18 May 1988)
Message-ID: <4275@medusa.cs.purdue.edu>
Date: 3 Jun 88 05:40:37 GMT
Article-I.D.: medusa.4275
Posted: Fri Jun  3 01:40:37 1988
References: <4245@medusa.cs.purdue.edu> <7348@swan.ulowell.edu>
Sender: news@cs.purdue.EDU
Reply-To: spaf@cs.purdue.edu (Gene Spafford)
Organization: Department of Computer Science, Purdue University
Lines: 60

In article <7348@swan.ulowell.edu> page@swan.ulowell.edu (Bob Page) writes:
>Isn't there a new backbone in Spain, fed by mcvax?

If so, no one has told me about it to update the map.

>I don't understand some of the map.  Like, why are ulysses, ubc-cs and
>vucomp on there, all dead ends (and vucomp a restricted link at that)?

Each is the gateway or main feed into an administrative or geographically
isolated domain.  ulysses is the feed into Bell Labs, ubc-cs is the
feed in British Columbia, and vuwcomp (not vucomp) is the gateway
into New Zealand.

>I submit that a 'backbone' link is to carry the news fast between
>other backbone links, and I don't see these sites as doing that
>(except for their leaf nodes ... hell, everybody does that). By the
>same token, alberta would have to be looked at.

Alberta is also the main feed into (surprise, Alberta).
Your definition of backbone is not the complete one.  Read the
text at the beginning of the article you are following-up to.

>I can't decide what the rutgers-ukma link is.  Two paths?  Does the '+'
>mean all the connected sites talk to each other?  Although it looks like
>rutgers is the center of the Usenet universe, some links are triads
>(rutgers/gatech/mcnc, rutgers/cmcl2/husc6, rutgers/ukma/gatech).

The two lines were result of a typo.  There is one link (NNTP)
between rutgers and ukma.  The + means a multi-way connection
(actually, all pairs meeting there are connected via NNTP links).
Can you think of a better way to show the map?

>Since I assume this is all NNTP traffic, I wonder how much Internet
>traffic would be cut by dropping some of those links, or making them
>mostly unidirectional?

Yes, all are Internet NNTP links.  However, many have been optimized
to be short hops.  Plus, the NNTP algorithm means that the full
article text is not shipped over ever link.  Once a site has it,
the total overhead to check if it arrived is only the article ID
plus a NACK packet -- not a lot of traffic, actually.

>I also wonder about linus ... we're a hop away from husc6, and I don't
>see much traffic through linus.

That doesn't mean the link doesn't work, or that there isn't much
flow.  It just means the connections to husc6 get called more often
and the news flows through there first.  If husc6 were down, you
would likely see a lot of news through linus as it took over the
slack.  

>Are these maps generated by hand?

Yup.

-- 
Gene Spafford
NSF/Purdue/U of Florida  Software Engineering Research Center,
Dept. of Computer Sciences, Purdue University, W. Lafayette IN 47907-2004
Internet:  spaf@cs.purdue.edu	uucp:	...!{decwrl,gatech,ucbvax}!purdue!spaf