Fragid semantics and language conneg

Section 3.2.2 of AWWW says that inconsistent fragid semantics is a
server error.

   representation providers must not use content negotiation to serve
representation formats that have inconsistent fragment identifier
semantics

This is in the context of a discussion of representations having
different media types, but the rule would also apply in the case where
there is only one media type involved, say text/html. RFC 2854 says:

   For documents labeled as text/html, the fragment identifier
   designates the correspondingly named element; any element may be
   named with the "id" attribute, and A, APPLET, FRAME, IFRAME, IMG and
   MAP elements may be named with a "name" attribute.

If one text/html representation gives one element, and a different
text/html representation gives a different element (perhaps in a
different language, or, as in the memento protocol, from a different
time), that sounds like an inconsistency to me - e.g. the two elements
might be of different types, or have different content, and a single
element cannot have two types, or more than one content.

Of course there is no problem here *in practice*. The fragid is doing
exactly what it's supposed to do - it's getting viewers to the right
part of the document, in the language of their choice. We just don't
have a theory that explains this.

The media types talk about what the fragid designates *according to
that representation*. AWWW (and RFC 3986) wants to talk about what the
uriref "identifies", independent of (but consistent with) all
representations. Maybe media type "designation" has to be
representation-relative, and different from "identification", which is
not. Or maybe "designation" is not connected with "semantics" (ugh).
Or else "designation" inconsistencies are fine, but lead to situations
where urirefs don't "identify" anything in particular.

If we have representation-relativity among representations for one
media type, why not relativity between representations having
different media types, or even within a single representation (per our
discussion of application/rdf+xml)? What consistency condition, if
any, is the most useful one?

I can't get too worked up about this but I don't like muddles. It
would be nice to have a straight story before issuing opinions on
things like 3023bis and fragids in redirects... can anyone think of a
fix? Looks to me - just off the cuff mind you - like the most
parsimonious fix is to say that fragids "identify" information
resources, abstracted over representation, with "representations"
defined in most cases according to the fragid sections of media type
specs - that is, reduce it to the previously unsolved problem "what is
an information resource". E.g. if GET http://example.com/doc yields a
representation with fragid aa "designating" an HTML element <ee/>,
then element <ee/> is a "representation" of the secondary resource
http://example.com/doc#aa .

Jonathan

Received on Monday, 1 November 2010 13:28:36 UTC