Re: Concern: XQuery <> valid XML

Jim Davis <jrd3@alum.mit.edu> wrote:
> At 02:42 PM 2/27/01 -0500, Michael Kifer wrote:
> 
> >Jeff Chapman <Jeff.Chapman@pervasive.com> writes:
> > >.. I
> > > would like to be able to build XML-based logic that can dissect and 
> > assemble
> > > XQuery expressions using normal XML tools/logic...
> >
> >If you want to construct queries on the fly, then the extraneous XML tags
> >which you have to (gratuitously) insert are an obstacle, not help.
> >People have been creating SQL queries on the fly easily for years (not that
> >SQL is a good language, but it would look even worse, if expressed in XML :-).
> 
> Michael,  I might be willing to concede that the XML syntax does not help 
> much in construction of queries (string appending is easy), and I might 
> even consider it a hindrance (albeit minor), but parsing XML queries is 
> undoubtedly made more difficult by the non-XML syntax.

Jim,
are you referring to the speed of parsing or to the difficulty of parsing.
I concede that XML makes things easier to tokenize, so there might be some
very small improvement in the speed of parsing, and it is slightly easier
to write a parser for it. However, I think neither point is significant
because the speed improvements aren't likely to be noticable and because
there will be off-the-shelf parsers that everyone will use.

This is why I fail to understand all these arguments about the need to
XML-ize the syntax of everything. In my view, XML is a data exchange
format, and standardizing is a good thing, because you need to query and
possibly reason about that data. However, exchanging queries is a different
matter.  A user doesn't write a parser every week --- he would simply
invoke one that is provided --- and a user application doesn't reason about
queries --- query optimizers do (and users don't write them).  So, I fail
to see the rationale for the XML-ization requirement for the XML query
languages.

> Moreover, requirement 3.2.1 says "One query language syntax MUST be 
> expressed in XML in a way that reflects the underlying structure of the 
> query. "  This requirement seems to me to be unmet by the syntax as proposed.

Yes. And as I said, I would like to see better explanations for the
rationale behind this. (I am not sure if this is the right forum, but I
think this issue might be of interest to many subscribers of this list). A
number of previous posters made references to visualisation, meta-level
assertions, tool building simplicity, etc., but I fail to see what any of
these have to do with the syntax of a query language and how XML-izing a
query language might simplify any of these things.



	--michael

Received on Wednesday, 28 February 2001 14:03:15 UTC