Re: URN single or multiple variants

JWilliams.sbd-e@rx.xerox.com
Thu, 16 Sep 1993 11:21:27 PDT

Date: 	Thu, 16 Sep 1993 11:21:27 PDT
From: JWilliams.sbd-e@rx.xerox.com
Subject: Re: URN single or multiple variants
In-Reply-To: "uri-request@mocha.bunyip:com:Xerox's message of 16 Sep 93 16:31:53 +0100 (Thursday)"
To: clw@merit.edu
Message-Id: <"16-Sep-93 19:21:27 +1".*.John_G._Williams.sbd-e@rx.Xerox.com>

the proposed model allows considerable latitude over what a URN references.
This is not a critism - but an observation.

Any cache machinery needs to know whether the object is mutable or not (and if
mutable how much longer the current instance might be expected to remain
valid). In the absence of any such information you can only assume the material
is mutable, and could change at any moment. As such caches are of limited use
in the absence of such information.

Users need to know something about the choices when derefercing (if details
such as the particaular location chosen to fetch from and the format are
determined at this point rather than being bound to a particular URN).

The user only has a finite set of software applications and convertors. Various
renditions may have different qualities (eg: resolution, a b/w derivative of a
color original). These discrepancies become particularly acute when dealing
with processible document formats where loss can drastically devalue a document
(eg: ascii is a poor alternative to MS-Word with diagrams etc,..). Then there
may be charges, and security, and other considerations too.

ques: what additional data do you propose to have (optionally?) available in
the dereferencing stage. It doesnt all have to be in place to have a workable
design - but if for example you want to support cacheing from day 1.....

ques: do you believe this design scales adequately. URN -> URL mapping implies
a directory lookup. If you think of the number of URNs you can expect in the
global internet if URN based systems are successful, then this is a completely
different order of magnitude to the problems existing directory services are
attempting to manage. For example do you assume the registration authority has
a map that knows the location of all copies of that named resource on the net?
I can see logistic and political problems here.

-- Jock Williams --