Date: Wed, 26 May 93 10:11:30 -0400
From: "Chris Weider" <clw@merit.edu>
Message-Id: <9305261411.AA06865@merit.edu>
To: marca@ncsa.uiuc.edu, peterd@bunyip.com
Subject: Re: URLs, URIs, and references
Peter,
I think you've made a fantastic exposition of the basic state of the art in
URLs. Also, I think that the current URN/URL architecture is exactly the
right one (although we are still arguing over, essentially, syntax). So, in
that spirit, My list is:
top: URLs must be easily extensible to new systems.
This is crucial, especially since we are getting a plethora of new
| systems.
|
| URLs are to be easily transmitted through existing infrastructure.
v Again crucial.
URL -> Metainformation mapping.
Crucial in my model, where all the typing of resources, format
negotiation, etc, is done with the URL and not with the URN.
URLs are to be easily transcribable without error (by machines).
Not so crucial, only the implementaors care whether it's easy or
hard.
URLs are to be easily manipulated by machines.
URLs are to be easily transcribable without error (by humans)
URLs should provide hints to the system about whether they may or
may not have gone out of date.
(This could be a potential nightmare, especially updating!)
bottom: URLs should provide strong typing information.
In my model that is provided by the Retrieve_Metainformation.
Since the URL is intended to be transient, I think that strong
typing, while quite desirable prima facie, could cause a lot of
things to break later.
URLs derivable from first principles...
is this to make creation of URLs easier? I don't see the utility
of this one...
URLs issued only by a specific server.
I don't think this is reasonable.
URL -> URN mapping.
I see this as a resource location service function. It could be
quite expensive, and of marginal utility. Do people primarily
hand around just URLs, just URNs or a URN/URL pair?
URLs on bar napkins.
This I think is most crucial, as technical progress in the
Western world would halt without this traditional method of
information transfer :^).
Well, there you go, Peter!
Chris