NetNews Usenet Archive 1992 #31/NN_1992_31.iso/spool/comp/mail/uucp/2490

Internet Message Format | 1993-01-02 | 3.9 KB

Xref: sparky comp.mail.uucp:2490 comp.bbs.waffle:5792
Path: sparky!uunet!cis.ohio-state.edu!zaphod.mps.ohio-state.edu!rpi!batcomputer!munnari.oz.au!ariel.ucs.unimelb.EDU.AU!werple.apana.org.au!tuple.apana.org.au!boombox!djk
From: [e-mail] (David Keegel)
Newsgroups: comp.mail.uucp,comp.bbs.waffle
Subject: Re: The UUCP network hasn't a tree structure
Message-ID: <[e-mail]>
Date: Sat, 2 Jan 93 09:09:01 +1100
References: <[e-mail]> <[e-mail]>
Reply-To: [e-mail]
Followup-To: comp.mail.uucp
Organization: Private site, Melbourne, Australia.
Lines: 66

In <[e-mail]> [e-mail] (Willard Dawson) writes:
>[e-mail] (Bruce Lilly) writes:

>> In article <[e-mail]> [e-mail] wr
>> >If they don't recognise it, they do the same (as long as there is a
>> >"tree" structure, and each node knows all the sites "below" it).
>>
>> What makes you think the UUCP network has a tree structure?

I'm really only going from my local experience.  Anyway, I *don't* think
the whole "UUCP network" has a tree structure.  The question is whether
a local section of it has some kind of tree structure (mine does).

The piece of it I am in (APANA) has about 50 UUCP nodes, with a single
connection to the Internet.  My machine is boombox.  It's only neighbor
is tuple, so I call tuple boombox's "feed" (probably an abuse of
terminology, but I know what I mean).

Tuple has about five UUCP neighbors, mostly "leaf" nodes like mine
(although one can also be a Fidonet<->UUCP gateway).  One of tuple's
neighbors is werple, which has about 20 neighbors.

About 5 of them are long distance calls (changed per minute rather
than the flat rate ($0.25) which Telecom Australia charges for local
calls (eg: same city).  I say werple is tuple's "feed".

Werple also has a SLIP link (SMTP packets only) to an Internet host,
and an NNTP-only link to another Internet host.

The APANA nodes outside Melbourne (where boombox, tuple, werple and the
two Internet hosts are) seem to have a similar topology.  One node in a
city dials werple and the others connect to that one.

The upshot of all this is that all APANA nodes (to my knowledge) have a
unique path from the host which is the destination MX for *.apana.org.au
and use the same path (backwards) to send traffic out of APANA.

This means that I can set boombox's smarthost to "tuple" as long as
boombox recognises mail for itself.  Tuple can set its smarthost to
werple, as long as it can recognise its "downstream" sites (one of
which is boombox, I think there is only one which is not directly
connected to tuple).

Being the de facto "hub" for APANA (through which all inter-city APANA
mail is carried for instance), werple needs a complete UUCP map of
APANA nodes.  But werple can send any mail for outside apana to The
Internet host it connects to (and thus to the outside world).

>I can't speak for other parts of the UUCP network, but my own little
>piece of it is an undirected graph.  That is, each link is bi-directional.
>And, many of the several sites with which I net have links between one
>another, each of which is itself bi-directional.

That isn't what I meant.  From your signature, you seem to have two
independant UUCP paths from/to The Internet.  If so, you certainly
don't have a tree structure (else it would be a tree with two roots).

I suspect this nice simple topology we have in APANA doesn't match
very well the normal practice in the USA and/or elsewhere.  I happen
to think this "tree" style topography is good for 20-100 UUCP nodes,
but perhaps this is skewed by having a flat-rate local phone charge.
(And APANA being a generally "low cost" network!)

--
<[e-mail]>  <werple!tuple!boombox!djk>  Tel: [phone]
aka: [e-mail], [e-mail].  Formerly: [e-mail].