RE: Are we losing the plot?

> > URIs are protocol elements intended for communication.
> 
> Really? Did the authors of RFC 2396 (and RFC 2396 bis) intend 
> that reading?

Speaking for myself: certainly. As for the others, the
context is clear:

RFC 2396 is a standards-track document in the IETF
(Draft Standard), and the intent for RFC 2396 BIS is for
it to become "Standard".  

RFC 2026 "The Internet Standards Process -- Version 3"
describes the scope of such documents. Section 1.1 and
3.1 ("Technical Specifications") apply.
 
> > To be useful, the definition of the URI scheme should
> > tell the receiver of a URI what resource the URI identifies.
> 
> What does "useful" mean in this context? I also really don't 
> understand what the rest of this sentence means.

RFC 2718 ("Guidelines for new URL schemes", best current practice)
calls (section 2.3) for "Demonstrated utility". I admit
that the text there is sparse. RFC 2026 generally calls
for specifications to be 'useful', and that the review process
includes some judgment. In this particular context, though,
a URI scheme definition is 'useful' if an implementor
of URI schemes (such as a browser author) can reliably
implement an appropriate action for their software
(such as a browser) given the URI scheme's definition.

> > When you define a URI scheme, you are expected to define the
> > access semantics of the scheme -- how it is that a receiver
> > of a URI in the scheme is supposed determine the resource
> > that the URI identifies.
> 
> Surely a reciever can only be dereferenced to a representation - and
not to
> a resource? According to RFC 2396 a resource is any thing that can be
> referenced by an identifier. 

For 'mailto:' URIs, there is no 'representation' to
'dereference'. I admit that the definition for
'resource' in RFC 2396 is weak, because the paragraph
given as a definition doesn't actually tell you what
a resource _is_, but dances around the subject.


> I guess I need to ask the question, what is the sphere of application
for
> URI? Is it for the Web as a discrete hypermedia application where URIs
are
> strictly /document/ identifiers?

Certainly URIs are not document identifiers when they are
'rtsp' or 'mailto' or 'telnet'. 

>                              Or is it for the Web as the global
> infospace, where URIs are used for making assertions about identified
> constructs.

I'm having trouble finding anywhere in RFC 2396 or the record
of URI development any references to this being within
the scope of applicability. The task of defining
identifiers that are useful for 'making assertions'
is definitely a hard problem in linguistics, AI,
and philosophy of language. 

There may be those who wish URIs would solve this
difficult problem.

>  Since URI is widely perceived as the "XML" (lingua franca) for
> identification systems, in providing a common syntax, 
> semantics and language for discussing identifier-related constructs,
> it is vital that the scope of URIs is properly articulated.

It is the sad plight of linguae francae to imperfectly
communicate. Yes, URIs are widely seen as solving many
problems, but it doesn't mean that they can.
(Note that I am overly pleased with 'tdb' in 
http://larry.masinter.net/duri.html as a way of
actually allowing URIs to identify any resource,
because all problems can be solved with another
level of indirection.)

> There's a clear need at this time to scope the sphere of 
> application for URI before we can even begin to propose
> new schemes for registration within the URI allocation.

Well, the scope is defined in the RFCs; would you
like to propose revised wording to that in
http://gbiv.com/protocols/uri/rev-2002/rfc2396bis.html ?

> Ad hoc revisionism is not helpful to anyone and can only 
> stifle development of applications that will contribute
> to the semantic web.

I'm not sure if this is intended as a complaint.

I hope I've provided adequate reference. Mail
archives for some of the URI mailing lists are
available at 
http://lists.w3.org/Archives/Public/uri/
going back to 1994, and Googling turns up earlier
notes (I haven't yet found a copy of the older
mail archives), but I've found

http://www.w3.org/Conferences/IETF92/LivingDocuments.html
http://www.cs-ipv6.lancs.ac.uk/ipv6/documents/standards/general-comms/ie
tf/uri/uri-minutes-92nov.txt
http://www.ripe.net/ripe/mail-archives/ripe-list/1992/msg00116.html

> We must collectively agree the scope of URI if
> it is to have any meaning and coherency.

Since the topic's been discussed for 11 years and
all we have are these poor RFCs to show for it,
I'm not confident that it will be possible to come
to a tighter consensus than we already have.

Larry
-- 
http://larry.masinter.net

Received on Friday, 19 September 2003 04:58:38 UTC