Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84; site cbosgd.UUCP
Path: utzoo!watmath!clyde!cbosgd!mark
From: mark@cbosgd.UUCP (Mark Horton)
Newsgroups: net.mail
Subject: Re: Want all paths from uucp -> ARPA
Message-ID: <1004@cbosgd.UUCP>
Date: Wed, 27-Mar-85 14:45:33 EST
Article-I.D.: cbosgd.1004
Posted: Wed Mar 27 14:45:33 1985
Date-Received: Thu, 28-Mar-85 03:18:40 EST
References: <1044@topaz.ARPA> <1184@houxm.UUCP>
Organization: Bell Labs, Columbus
Lines: 96

I am not a "person of authority" on the ARPANET, but I can shed some light.

The official ARPANET policy is that anyone who causes a packet to be
sent over the ARPANET is "using the ARPANET" and better be doing so
only to fulfill a DOD contract.  Any other use of the ARPANET is illegal.

While there has never been an official statement of what this means for
mail bridges (sometimes incorrectly called "gateways"), most people in
the know think that if an official interpretation were to be asked for
the interpretation would be "if a person on UUCP send a message, via
a bridge, to a person on the ARPANET, and that UUCP person has not
been officially granted permission to use the ARPANET, this person is
using the ARPANET, and the bridge is violating ARPANET policy by serving
as an unauthorized gateway."

If this issue were to be pursued, it is possible that the sites serving
as bridges (as a courtesy to both communities) might be pressured into
terminating the bridge service.  So we have not pursued it.

Practical reality is another matter.  Even within the ARPANET itself,
there has always been lots of non DOD related traffic.  Any university
that is on the ARPANET views the entire university (or at least the
parts of it that the ARPA contact trusts) is "on the net", and allowed
to use it for whatever they like.  There is lots of traffic by legitimate
researchers talking to other legitimate researchers about research they
are doing that is not funded by DARPA.  There is lots of personal traffic.
(Not unlike UUCP.)  There are even big mailing lists, like SF-LOVERS,
that clearly are not DOD related.  It got to the point where people would
list their ARPANET mailing address in their publications and talks, since
they knew that most people who were doing related work were also on the
ARPANET.  (Look in any issue of SIGART, for example.)

Now that TCP/IP is the norm, there are lots of local nets at the various
universities.  People on the local net are also allowed to use the ARPANET
as though they were directly on it, but they are not on it and are subject
only to the requirements of the local net itself.

I have made the analogy that if a DOD person has a terminal on the ARPANET
in his office, it's not unlike having a telephone or a paper mailbox on
his desk.  If I want to send this person paper mail, I'll send it via
US mail and the local delivery person will carry it down the hall and
drop it in the mailbox.  I don't get arrested because I caused this
delivery person (a DOD resource) to carry my letter down the hall.  Ditto
for the phone, assuming the PBX system the DOD organization has is
reachable via the public network.  I don't see why electronic mail should
be treated any differently.

Now, we look at the current situation, as it has evolved in practice.
The UUCP community has grown to the point where it is larger than the
ARPA Internet.  There are a few mail bridges who are willing to pass
mail between the two (without official permission.)  People depend one
these bridges to get their jobs done.  It doesn't cause anybody any
problems to forward mail from one net to the other.  In fact, it's more
likely to cause a problem when an ARPA person sends mail to UUCP, since
they caused some UUCP machine to place a phone call and run up their
phone bill.  The ARPANET uses leased lines with a fixed cost, so additional
traffic (as long as it doesn't overload the network) doesn't cost anything
extra.

I am told that the people who run the ARPANET are aware of UUCP and other
networks, and are not unhappy with the status quo.  They realize that
people are getting useful work done and do not want to clamp down.  They
are worried, however, that someone might someday do something nasty to
some machine on the ARPANET.  They want to be able to invoke their official
policy if this happens, shutting off the bridge and preventing such access
from happening again.

The one major thing that ARPA does not want to have happen is the rest of
the world using them as a forwarding mechanism or transport mechanism
between two outside networks.  If I am on UUCP and I want to send mail to
CSNET, right now there is no CSNET-UUCP mail bridge (that I know of) so
I have to send my mail via either the ARPANET or BITNET.  If I send it via
the ARPANET, I'm violating both official policy and unofficial policy,
e.g. this is a Bad Thing.  If I'm in California and I place a local phone
call to my nearby ARPA-UUCP bridge, which sends the mail to New York,
which places a local call to the destination, what I've basically done is
to use the ARPANET to avoid a phone bill.  I am told that this is not
supposed to be allowed to happen.

Since some networks only seem to be attached to the ARPANET, there is no
other way to get to people on them, and I don't feel guilty about going
via the ARPANET.  This especially applies to LAN's at universities on the
ARPANET, but also applies to MailNet.  In practice, it may apply to
CSNET as well, I don't know of a way to get to them except over the ARPANET.
(I would prefer, however, to see CSNET-UUCP bridges set up.  Since there
are UNIX machines on both, there is no reason for this not to happen.)

However, it is bad form to use the ARPANET to go between UUCP and BITNET
(there is an official gateway at Penn State) or UUCP and DEC (there is
a gateway at DECWRL) or worse yet, between UUCP and UUCP.

My reading says that as long as one of the two people involved in the
mail transaction is on the ARPANET, it's OK.  However, I must disclaim
authority on this - I cannot speak for the ARPANET.

	Mark Horton