W3C home > Mailing lists > Public > uri@w3.org > September 2001

Re: Excess URI schemes considered harmful

From: Rob Lanphier <robla@real.com>
Date: Wed, 26 Sep 2001 09:49:32 -0700 (PDT)
To: Mark Baker <distobj@acm.org>
cc: <michael@neonym.net>, <uri@w3.org>
Message-ID: <Pine.LNX.4.33.0109260941070.11712-100000@mmmm.robla.org>
On Wed, 26 Sep 2001, Mark Baker wrote:
> Oops, sorry, missed this one.
>
> > On Tue, Sep 25, 2001 at 02:52:50PM -0400, Mark Baker wrote:
> > > There's different degrees of equivalency at play here.
> > >
> > > It is the case that if two HTTP URIs are equal, that they necessarily identify
> > > the same resource (also see RFC 2616 sec 3.2.3).  It is not the case that if
> > > two HTTP URIs are *not* equivalent (even modulo 2616/3.2.3), that they do
> > > *not* represent the same resource.  The only way to determine whether two
> > > non equivalent HTTP URI represent the same resource latter is to have
> > > authoritative info about that equivalence (or lack thereof).  One way for
> > > that authoritative info to be communicated is with an HTTP redirect.
> > >
> > > Even the current CTURI draft has the latter problem, as the encoding is not
> > > canonical.
> >
> > This is indeed the case with _all_ URIs. Equivalence of resources across
> > schemes is iff the URIs are lexicographically equal.
>
> Untrue.  See below.
>
> > URIs themselves
> > express no other type of equivalence other than that since equivalence
> > is always application specific and more often than not is also context
> > specific. Unless you have some application specific knowledge talking
> > about equivalence of Resources is a very dangerous and, IMNSHO, simply
> > a Bad Thing To Do...
>
> Luckily, HTTP is an application protocol, so is in a very good position
> to define application specific knowledge/context, such as when http
> URIs are "equivalent" or not.  Non-http URI schemes don't have that
> advantage.
>
> IMO, yet another reason why using http URIs is Goodness.

Are you saying that HTTP is the only protocol/scheme allowed to define
equivalencies?  Wrong.

A "ContentType:" scheme can define equivalencies.  Moreover, it can define
equivalencies that map directly to equivalencies that already exist in the
822 header world.

If the CTURI proposal doesn't define a way of mapping to a canonical form
for comparison purposes, then I would agree that should be addressed.

Rob
Received on Wednesday, 26 September 2001 12:50:18 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 13 January 2011 12:15:29 GMT