W3C home > Mailing lists > Public > uri@w3.org > October 2003

Re: Announcement: The "info" URI Scheme

From: Roy T. Fielding <fielding@apache.org>
Date: Thu, 2 Oct 2003 15:15:05 -0700
Cc: uri@w3.org
To: "Daniel R. Tobias" <dan@tobias.name>
Message-Id: <E06DE134-F525-11D7-BD07-000393753936@apache.org>

> But if, instead, a non-HTTP URI is used as the "canonical" means of
> referring to books by ISBN (as, in fact, the "urn:isbn" namespace has
> actually been registered and has an RFC defining it, as I recall),
> then the end user has more flexibility in deciding what information
> he/she wishes to retrieve on it.  When typed into a browser or
> followed as a hyperlink, a "http" URI will always cause the resource
> set up by the "owner" of that URI to be retrieved (or failed to be
> retrieved in the case of 404 or DNS errors),

No, it doesn't.  I recommend investing a little time figuring out the
numerous ways in which browsers can be configured to use proxies, not
just by scheme but by entire URI prefix, via both manual and automatic
configuration methods.

> while another scheme is
> capable of user-driven configurability with regard to how to treat an
> attempt to dereference it (or would be if browsers were sufficiently
> advanced to give this capability, as I'd hope they'd be if such URIs
> were in wide use).  I'd be able to set my browser to go to a page
> related to that ISBN at niso.org, or amazon.com, or some other site
> of my own choosing, or make a database query, or bring up a local
> file from my own system, or whatever else I chose.

Which is exactly what it can do right now with "http" schemes.

Let me make this clear, since some people didn't understand my
last message about "http" schemes:  I specifically oppose the use of
"http" identifiers to identify everything, since there really is no
technological reason to prefer one widely deployed scheme over others.
However, I also don't think b.s. arguments about the limitations of
"http" or the invulnerability of non-http schemes to the same issues
that "http" faces should be left unanswered.

There is no good reason to have an info scheme. There are good reasons
to have properly designed schemes for Dewey Decimal Classifications
(hierarchical) and US Library of Congress Control Numbers (flat),
and there is at least some justification to use the "urn" scheme
for the simple reason that one uber-scheme is more than enough.

I have no problem whatsoever with NISO doing the minimum effort
needed to define those (and all the other) namespace authorities by
means of the RFC process, or editing the URI scheme registration
process so that it isn't so apparently difficult to register new
schemes.  However, I will note that the main reason there aren't
more registered schemes is because authors foolishly believe that it
is easier to register one uber-scheme than it is to simply register
all of the meaningful naming mechanisms or authorities that might
reside within such a scheme.  Experience has shown that to be false.

....Roy
Received on Thursday, 2 October 2003 18:15:26 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:25:06 UTC