Hi, Tony.  You've had a lot of responses already, so I'll just respond
to the resonses rather than your original question.

Tony Hammond writes:
 > I guess if that is the real position then we could support SRU URL
 > in and SRU XML out for power users, and SRU in and arbitrary XHTML
 > out for regular end users in an MGX sort of way.

Others have made alluded to this point but it's worth making
explicitly: the real distinction is not between power users and
regular users; it's between humans (who read the results) and program
(which have to analyse them).  Leaving aside matters of what is and
isn't SRU-conformant, if you just want to make your search results
visible to eyeballs, then HTML (including XHTML) is appropriate; but
if you want it to be amenable to further processing -- for example, by
metasearch engines -- then you need to return compliant XML.

 > Wouldn't that mean we could offer full SRU compliance for those who
 > could consume the SRU XML, and at least standardize the request
 > half on an SRU querystring?

Yes.  Supporting CQL for queries is better than not doing so; but as
Ralph's already said, it is the easier half of the search-and-retrieve
equation.  The big wins come from standard result formats.  (That's
why there are so many more initiatives that deal with metadata
representation than with query formulation.)

Eliot Christian writes:
 > It seems equally valid to me that people have SRU layered
 > underneath an intermediary that renders HTML. I don't see why it
 > should matter to the SRU community whether that intermediary has
 > been written with the stylesheet programming language or any other
 > programming language.

I don't think anyone does care _how_ XML is transformed into HTML.
The question is whether the raw XML is available to clients _at all_.
If it isn't, then the service being offered is not SRU (nor MXG).  If
the XML is avalable, then, sure, it's nice also to offer a service
that can transform it from this machine-readable form into something
human-readable.

Oh, and while I'm here, Ralph LeVan writes:
 > You can use CQL as the query grammar embedded inside your own
 > unique URL's and return arbitrary XHMTL.  But CQL wasn't designed
 > to be a human entered query grammar.  It does simple stuff well,
 > but tricky stuff gets tricky.

Tricky stuff is tricky in any language.  The one area in which CQL
fails as a human-usable language is the perverse interpretation of
queries such as
	tottenham court road
in which "court" is interpreted as an unsupported relation.

 _/|_	 ___________________________________________________________________
/o ) \/  Mike Taylor  <[log in to unmask]>  http://www.miketaylor.org.uk
)_v__/\  The problem with democracy is that it's government of the people,
	 by the people.  Have you _seen_ the people lately?