UR accessors

Ronald E. Daniel (rdaniel@acl.lanl.gov)
Mon, 3 Jul 1995 13:52:13 -0600

From: "Ronald E. Daniel" <rdaniel@acl.lanl.gov>
Date: Mon, 3 Jul 1995 13:52:13 -0600
Message-Id: <199507031952.NAA17458@idaknow.acl.lanl.gov>
To: uri@bunyip.com
Subject: UR accessors

Hi all,

I had meant to respond quite awhile ago to Dan's message about UR
Accessors. However, that reply got lost in the shuffle in my quest for
an adequate mailer. So, here is a short version. 

First, I agree with Leslie that URAgents can legitimately serve as an
identifer. To be sure they are a more abstract form of identifer than a
simple name, but they can identify resources in a very useful fashion.

Dan's message was triggered by the concept of a URC0 - which was the
notion of two resources being equivalent but not bitwise identical. As
one on the authors of the urc0 draft, let me state that we never
intended to promulgate such a concept. The text/x-urc0 format was
proposed so that the URN testbed people would have a URC fomrmat that
was very simple to parse while they worked on other parts of the problem.
URLs could have metadata following them so they could be distinguished on
the basis of format, price, etc. The only reason that such information
was not made more explicit was that we thought people wouldn't really
care to distinguish between different URLs in the URN testbed.

Dan's questions about the URC0 concept seem to boil down to equivalence
relations; which ones should we specifically call out and use. I think
we already have a couple of very useful relations defined. The first is
that things named with the same URN are stated, by the naming
authority, to be equivalent in some sense. That sense may vary from one
naming authority to another, but that is the defining characteristic of
a URN. Second, we have a stronger equivalance relation - bitwise
equality. This is exactly Keith Moore's notion of LIFNs, and we do not
need a new term for it. So, the question now seems to be "are there other
equivalance relations we should identify and utilize"? There do not seem
to be many useful, stable, categorizations whose strength is intermediate
between the URN and LIFN cases. Same version, differenc formats is about
the only one that I can think of. There are many useful relations that
are looser than URN equality. Equality of subject,
equality of author, equality of publication date, etc.

While these looser, dynamically determined, relations are useful, a good
name for them eleudes me. For some reason, UR INformation Equivalances
(URINE) just doesn't seem appropriate. (Apologies for the lavatorial
humor - its this upcoming trip back to the UK, I swear. :-)

Name aside, they do seem to be the sort of equivalances that can be
determined by URAs. I recommend we continue to use the acronym URA for
Agents, have it denote an executable specification, and continue to
discuss them in the URI-WG. Perhaps someone else out there can take
this notion of equivalance relations and push it further.