[net.mail] Uniqueness of Names

davecl@shark.UUCP (11/15/83)

You want to have names in the same "domain" unique.  You do not want to
have to force the entire universe of names to be unique.

While the syntax is poor, remember that the current uucp source routed
syntax, and the practise of giving your "address" relative to a well known
"host" has in fact already provided us with a domain facility similar
to what the ARPAnet is currently experimenting with.

In particular, there is nothing wrong with the name "sun", relative to
the Sun Corp., and the name "sun", relative to the University of Maryland
meaning two completely different systems.

Dave Clemans
Tektronix

dw@rocksvax.UUCP (11/17/83)

I agree with Dave Clemans that names must only be unique within
a domain.  The question then becomes one of deciding what constitutes
a domain.  Current discussion on the Arpanet seems to be towards
defining a single "uucp" domain.  However, I am beginning to wonder
whether this is the best scheme for us.  Perhaps what is needed
is separate domains, or perhaps a single top level domain for
uucp and a whole bunch of subdomains for different areas, companies,
etc.

My auto-routing software (which only looks at messages which orginate
locally on rocks34) believes that the names of uucp sites are unique.
I have not thought of a beter way to do it under the present situation
of no "officially" defined domains or sub-domains for uucp.

Don Wegeng

{rocks34, rocksvax} !dw
Wegeng.Henr@Parc-Maxc.ARPA

gnu@sun.UUCP (John Gilmore) (11/18/83)

	While the syntax is poor, remember that the current uucp source routed
	syntax, and the practise of giving your "address" relative to a well
	known "host" has in fact already provided us with a domain facility
	similar to what the ARPAnet is currently experimenting with.

That's a translation I'd never thought of:

	decvax!microsoft!sco!root
		--> root@sco.microsoft.decvax.uucp

It works!

This also points up a possible way around the "to route or not to route"
question for uucp sites.  While there is no guarantee that "sco" is The
Santa Cruz Operation, it IS true that "sco.microsoft.decvax.uucp" is them.
Since most horribly routed mail is news replies, it probably follows very
predictable paths.  We can convert a known horrible path to a known good
path, rather than translating site names out of context.  For example,
you might transform that path to "sco.ihnp4.uucp" since you could know that
the two were equivalent but this one is faster.

This domain syntax is even in the grand Arpa tradition of tying the domain
name to the physical link involved...

mark@cbosgd.UUCP (11/19/83)

With current UUCP software, it's important for all names to be unique,
because UUCP assumes it.  If you are one of two sites named "foo" and
you call site "bar" which believes "foo" is the other one, they might
send you mail destined for their "foo" which happens to be queued up.

However, in the long term, 7 letters is not enough space to keep all
the names unique.  Several cases of non-unique names have been cited
here - interestingly enough, the Marx brothers were the first such case:
Bell Labs in Whippany got them first, and the University of California
at Davis was amazed when I told them they couldn't have them.

There's a much worse case nobody has noticed: the trend to name computers
after letters of the alphabet.  Computer centers are especially good at
this one.  They always have the A machine, the B machine, and so on.
The good news is that this leaves 6 letters to name the computer center,
so we have houxa (BTL Holmdel), ihuxa (BTL Indian Hill), ucb-cfoa (Berkeley),
(this one is, of course, botched, but they don't run uucp so it doesn't
matter), and so on.

The rule we're going to eventually have to settle on will probably look
something like this:

Your real name is a fully qualified domain, e.g. d.osg.cb.btl.uucp (we
name our machines after letters of the alphabet too) and that real name
is required to be unique in the world.  You also have a UUCP name (at
least as long as we still run UUCP) such as cbosgd.  The UUCP name might
convey less information than the full domain name, but it has to be
unique in the first 7 characters.

Each level in the domain tree has to have a person who makes sure that all
names in that domain are unique within the domain.  The top level will make
sure there are no other UUCP domains.  There will be someone appointed
keeper-of-the-books for UUCP that will make sure that nobody picks a name
that's already in use.  BTL will have another such person.  And so on.

	Mark Horton