Re: URNs, handles, and Stockholm

William Y. Arms (warms@CNRI.Reston.VA.US)
Thu, 15 Jun 1995 16:52:36 -0500


Message-Id: <ac060484040210042294@[132.151.1.217]>
Date: Thu, 15 Jun 1995 16:52:36 -0500
To: Michael Mealling <Michael.Mealling@oit.gatech.edu>
From: "William Y. Arms" <warms@CNRI.Reston.VA.US>
Subject: Re: URNs, handles, and Stockholm
Cc: uri@bunyip.com, dely@CNRI.Reston.VA.US

Michael,

>If any offense was taken please let me know. I'll make sure to buy
>you a beer or two in Stockholm.

No offense was taken, but I'll split a couple of beers with you anyway.

To put things in context, apart from myself, we currently have 5 people
plus an intern working on the Library of Congress copyright registration
system, of which the handle system is a small but important part.  We could
easily justify 10 on handles alone, of whom half would be development and
half in support.

The following are some notes to myself that I had not planned to distribute
but they might help the discussion.

Bill

-----------------------------------------------------

My basic assumptions are:

1.  There will be more than one URN scheme.  The various schemes are
solving related but different problems.  For example, the philosophy behind
the handle system is that it is a pure naming scheme.  The system can be
used to identify any resource and makes no assumptions about the
applications and systems that those resources form part of.  (This is in
line with Larry Masinter's recent message recommending that the URN and URC
questions be kept separate.)

2.  A full system to maintain and resolve URNs is a big project.  (The
handle system will end up at about 200,000 lines of code.  Much of the code
lies in scale-related functions, such as administrative tools, security,
fault tolerant operation, load balancing, caching and proxy servers,  etc.)
Because of the size, only a few resolvers will be implemented in full and
supported.  Both DNS and the handle system are general purpose systems that
could be modified to support a variety of URN schemes.

3.  A URN name space can not be the same as the domain name space.  (I
completely agree with Mike Shapiro's message on this topic.)  Maintaining
the name space over time is another big task.

These assumptions lead to the suggestion that the IETF working group should
concentrate on identifying the isssues that concern all of the URN schemes,
rather than attempting to decide between schemes, I suggest that:

(a)  The working group should recommend for full field trials any scheme
that satisifies RFC 1737 or some other set or criteria to be agreed.

(b)  At Stockholm, the working group should endorse a recommended syntax.
Any reasonable syntax will do.  The recent draft by Ron Daniel and Paul
Hoffman seems close to the mark.

(c)  The working group should prepare a list of common issues that require
help from others.  In particular, there should be a standard client
interface so that a client (e.g., a Web browsers) needs only one set of
modifications to interface to all schemes.

(d)  We consider the issue of how to run a global naming authority.

WYA