Date: 05 Apr 94 08:17:02+0200
From: pays@faugeres.inria.fr
To: ccoprmm@oit.gatech.edu, moore@cs.utk.edu
Subject: Re: Seattle minutes
Message-Id: <765526622.24592.0-faugeres.inria.fr*@MHS>
I fully agree with keith comments.
Maybe I am especially dumb but from what I have read and heard in Seattle
I am unable to understand what people mean by URCand/or see how
they will be used.
Our own experiments with documents/resources names and locators
have shown a need for different things
1. attached to a name: a set of information (characteristics?)
which are bound to the document itself (whatever the instance,
the format, the encoding and the location) such as Title,
Authors, Keywords (or even abstract)
2. attached to an instance: another set of information
which are specific to that instance such as Format, Size,
location, and possibly acces rights, or access cost
I don't believe it is realistic to expect to store all this information
in a single entity.
Additionaly I strongly support the idea of discussing over the net
scenarios about how all this will be used. The keyword here is
to analyse and meet typical users needs.
This will certainly lead to define a conceptual model which is badly needed
and that we will be ready to experiment with.
a first unordered set of user requirements:
- from a document name: be presented the first set of info
(document descriptor) AND info about existing instances
- from an approximate document name AND a few chracteristics VALUES
be proposed a number of matching documents (name)
- from a locator be provided direct access to the resource
(eg. what happens today with WWW acces from locators)
- from a locator on explicit demand be provided with this
specific instance set of attached information (resource descriptor)
- from a locator on explicit demand be able to get acces to
the document name and thus descriptor
First questions that arise:
Are we satisfied with only a URN and a URL wich implies that
there exist an implicit functional binding between these
and the respective associated "descriptors" (eg. such as a suffix)
Do we prefer to have explicit entities (with independant names)
for these descriptors and thus to have to define the explicit
name binding between all these entities?
regards,
-- PAP