Re: Using fragment identifiers with URNs

On Fri, Sep 28, 2001 at 12:02:00PM +1200, Stephen Cranefield wrote:
> Roy Fielding wrote:
> > It is impossible to define a name that cannot be used for retrieval.
> > The act of identifying has no responsibility for how that identifier
> > is used.
> 
> No, but the act of defining an identifier scheme could be seen to have
> that responsibility.  It seems reasonable for a scheme to clearly define
> its intended use (although many schemes may make no such constraints).

Fragments aren't part of the URI standard thus, a given scheme
shouldn't be talking about fragments. The use or non-use of fragments
belongs in an application definition (along with how it uses URIs), not
in a specific URI scheme definition.

> > I doesn't matter what a given URN nid may claim about the 
> > properties of an identifier in its namespace, the act of assigning a
> > name provides the world with an identifier that might be used by some
> > system somewhere, through some unknown number of indirect or redirected
> > resolutions, to perform a retrieval of a representation of the concept
> > identified by that name.
> > 
> > It is therefore never correct to say that a fragment is 
> > scheme-dependent, even though there are some schemes that identify
> > resources for which no representation is ever suitable for fragment
> > references when those references are made within the Web system.
> 
> Thanks for that wonderfully clear and unambiguous statement of your
> view of the principles behind URIs.  Does that represent the shared view
> of the committee that created RFC2396, or is this your personal view?
> (I'm not trying to suggest that you aren't an authority here, I'm just
> trying to gauge whether or not I can consider the statements above
> as definitive.)

Being involved in the discussions around what became 2396, I can pretty
comfortably say that Roy's assertion above is consistent with what
everyone involved considers the role and scope of fragments to be.
It was painful getting there but once we did we realized it made good sense.

> I hadn't realised that RFC2396 was intended to say anything about what
> applications *might* choose to do with URIs outside Web architecture,
> but I see that is probably necessary so that URIs can be incorporated
> into future techologies.

2396 was attempting to say what URIs themselves were sans any application
context (be it hypermedia or directory services). The inclusion of the
URI References (and thus fragments) was an attempt to be clear about
one of the most common uses of URIs. The problem might be caused by
the definitions being used for the "web architecture". At the time 
the most common definition of the 'Web' 'the set of resources identified
by a URI'. If you're using a definition that assumes things like
URI References, caches, media types, etc then I can see where the
confusion would be...

-MM

-- 
--------------------------------------------------------------------------------
Michael Mealling	|      Vote Libertarian!       | urn:pin:1
michael@neonym.net      |                              | http://www.neonym.net

Received on Wednesday, 10 October 2001 14:35:04 UTC