Received: from CMU-CS-CAD by CMU-CS-A; 27 Oct 83 14:30:18 EDT
Date: Thursday, 27 October 1983 13:58:15 EDT
From: Robert.Frederking@CMU-CS-CAD
To: bovik@cmua
Subject: UUCPnet/Usenet: the true story
Message-ID: <1983.10.12.20.15.9.Robert.Frederking@CMU-CS-CAD>

Apparently UUCP net is the mail network, while Usenet is the news network
that uses UUCP (or something like that).  It is always flaky.  The slashing
convention mentioned in an earlier note in this message file is only needed
if you are typing at the Unix shell level, to quote the "!"s in the address.
If you are inside a program like Hg, you must *not* put in the slashes.  The
network consists of a set of dial-up links, each of which waits a different
amount of time to call the others, depending on whether it's got messages
queued up, with periods ranging up to once a week.  It tries to warn you
about non-delivery, but doesn't always succeed.  "uucp" is the "unix-to-unix
copy" command, and is the method by which Berkeley distributes its Unix
updates.  It has apparently spontaneously turned into a network, and has
*no* central administration!  (Libertarians should love it.)

The address is interpreted by using the ARPAnet part first, and then
interpreting the UUCP part left-to-right, so <h1>!<h2>!<h3>!<user>@<h0> goes
to hosts <h0> (on the ARPAnet: Berkeley, for example), <h1>, <h2>, and <h3>,
where <user> lives.  Thus routing is done by *you* (since none of the
machines know where all the others are).  The current issue of Byte (as of
10/12/83) has a map of Usenet.  The return address for the above would be
<h2>!<h1>!<h0>!<sender>@cmua.  Some nodes are very flaky; some are very
good.  If you know "good" paths through the net, you're better off.  If you
go through a "bad" node, you're messages may take forever to get through,
and may also disappear without a trace.  Craig Everhart's UUCP path-finding
program (on the A, RUN ROUTE[c410ce10]) unfortunately doesn't know which
nodes are flaky.

Apparently Berkeley is not actually a gateway, but only forwards certain
mailboxes.  There are a number of connections to ARPAnet, including our own
cmu-vlsi (this is how we get Unix updates, apparently).  One person said
that @uw-beaver works better than @berkeley; another said that @seismo works
well, is fast, and has many of the Usenet sites in its tables.  There is no
built-in limit to UUCP mail size, although Berkeley imposes a 100K limit,
and apparently a bug in the PDP-11 version may trash mail larger than 32K,
if the path goes through an 11.

Use ARPAnet if you can.