Message-Id: <9310170421.AA11945@tipper.oit.unc.edu>
To: "Rob Raisch, The Internet Company" <raisch@internet.com>
Subject: Re: URNs and Meta-Information - The Value of ISBN
In-Reply-To: Your message of "Sat, 16 Oct 93 17:19:45 PDT."
<Pine.3.03.9310161744.A11995-c100000@hmmm.internet.com>
Date: Sun, 17 Oct 93 00:21:22 -0400
From: Simon E Spero <ses@tipper.oit.unc.edu>
The reason for having multiple naming spaces for URNs is to allow existing
namespaces be converted to URNs without massive re-engineering. For example,
I might want to give a reference to "The SGML handbook". By enclosing a
URN of the form <urn://isbn/0-19-853737-9>, you could automatically
determine whether that book was the same one as was on your desk. In addition,
you could make use of special knoweldge of that higher level domain to
produce a better search strategy to locate an instance of this resource.
In the case of an ISBN, the best strategy might be to perform an ISBN
lookup using Z39.50 against your local library.
If we didn't have any domain specific information for this namespace, then
we could still resolve it using our default resource resolution system
(whois++, email to peterd@bunyip.com (works for hockey tickets), etc).
The only operation defined upon URNs is equality; however, this is all you
need to perform an associative lookup. Whilst I have some suggestions which
I'll present in Huston for certain extra properties to make URN resoultion
easier, specifying a resolver in the URN seems a little too much like cheating
to me :-) :-)
I'm not sure if I was the origin of the multple namespaces concept; they were
definitely in the fabled midnight to 3:00am mid-beer-bof draft I prepared
at the Washington IETF. One part has already disappeared (version numbering),
that wasn't a crippling blow, as you can always wrap things up another level.
If you remove multiple namespaces you can't do the same surgery. They're
gone for good.
If we must get rid of a field, I'd vote for eliminating a mandatory
publisher identifier. For almost all classes of namespace this field
is essential, but for several major classes, including ISBN and ISSN,
it's redundant. Remember - a null publisher ID would mean 5 (five) colons
in a row. Wouldn't that be really irrigating, I mean irritating.
Simon