W3C home > Mailing lists > Public > w3c-dist-auth@w3.org > October to December 2001

Re: RFC2518 (WebDAV) / RFC2396 (URI) inconsistency

From: Roy T. Fielding <fielding@ebuilt.com>
Date: Mon, 26 Nov 2001 15:59:23 -0800
To: Jim Whitehead <ejw@cse.ucsc.edu>
Cc: "'WebDAV'" <w3c-dist-auth@w3.org>, uri@w3.org
Message-ID: <20011126155923.A11448@waka.ebuilt.net>
> I agree that it makes sense to have a different identifier for each revision
> of a URI scheme. BUT, these identifiers do not identify the same resource as
> the scheme URI.
> 
> For example, "http:" identifies
> 
>     The abstract notion of locators for Web
>     resources accessed via the HTTP protocol

No, it does not!  The http scheme defines a namespace with a hierarchical
syntax for which the naming authorities are defined by DNS/IP address and
TCP port.  The protocol used to access those names will change over time,
just as it has already done so twice.  Whether or not all such protocols
are called HTTP depends on the IETF, which doesn't move anywhere near as
fast as we develop new protocols that use the http namespace.

> which is a different resource than
> 
>     The set of locators for Web resources
>     accessed via the HTTP protocol and defined
>     specifically by RFC 2616.
> 
> and different still from
> 
>     The set of locators for Web resources
>     accessed via the HTTP protocol and defined
>     specifically by RFC 2068.
> 
> even though the actual URL space in question could be the same (that is, the
> abstractions are different, even if, by chance, all these abstractions refer
> to the same set of URLs).
> 
> So, you could still have "http:" and two additional URI/URL/URNs for the
> version-specific definitions of the http URL space.

Yes, but only if we defined it to be such a thing, which we won't do because
Web browsers still discard the scheme name if it is the same as the scheme
of the referring document.

> >  But that is not just my preference -- it is also
> > demonstrated by that silly "about:" URI in Netscape Navigator.  That does
> > identify a resource, not the naming scheme.
> 
> Actually, since it does consistently return a resource, doesn't that make it
> a URL?

*shrug*  It isn't a relevant distinction.

> > In other words, I think that "scheme:" is only a valid identifier for the
> > namespace if the scheme defines it as such.
> 
> This would only be for pragmatic reasons, not for philosophic ones.

Hey, I'm a very pragmatic philosopher.  I don't care for philosophies
that don't work in practice.

....Roy
Received on Monday, 26 November 2001 19:04:17 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 2 June 2009 18:43:59 GMT