[
Lists Home |
Date Index |
Thread Index
]
At 09:18 AM 9/26/2002 +0100, Bill de hÓra wrote:
>Of course. The problem with languages and specifications conceived as
>though they were the technological analog of finger food is not choosing
>a subset (given a platter, any fool can pick and choose), but making
>sure all the subsets usefully intersect. If there is a useful (that is,
>interoperable) intersection to be found with such a specification, it
>often becomes the basis for a future specification, the analog of a
>square meal, whose creators will be praised for deciding what to leave
>off the plate. Never trust a silver platter.
You seem to be implying that XQuery is a really big language. It is
certainly bigger than Quilt was, mainly because of support for types. But
if you look at the grammar [1] the whole language is 89 productions, and
most of that is either the XML grammar, which you obviously need for
constructing XML, or the XPath grammar, which you obviously need for
finding things in documents. Add to that FLWR expressions, used for joins,
function definitions and function calls, and the ability to declare
namespaces and import schemas, which all seem obviously important, and the
rest is small change.
Except for the syntax for matching types, which is larger than I would
prefer (10 productions out of the total 89), largely because of the
structure of XML Schema's type system. I believe that there are clear uses
for strongly typed data [see 2], but not all users will want this, and not
all implementations will provide it. So I think that there will be one
community that uses strong typing, and another that does not. There will
also be some implementations that provide strong typing statically, and
others that provide it only dynamically. We have three different
conformance levels to accommodate these audiences [3], but two of them use
exactly the same language and simply have different requirements on error
reporting, and the other simply disallows schema imports or naming of types
that are not built-in to XML. If you leave out the syntax for matching
types, I believe that XQuery is roughly the same size as XSLT + XPath.
Incidentally, section 5 of the XQuery document [4] shows scenarios for
which people had asked for special syntax, demonstrating that these things
can be done without additional syntax because of the compositionality of
XQuery. We have always emphasized compositionality instead of adding new
syntax. I don't think that there is that much more that could logically
have been left off the plate, except for schema import and support for all
the types that enables, and we have a conformance level that does leave
that off.
We have people asking for features that we do not support and probably will
not support in XQuery 1.0. I think that's a good sign.
Jonathan
[1] http://www.w3.org/TR/xquery/#id-grammar
[2] http://www.w3.org/TR/xmlquery-use-cases/#strong
[3] http://www.w3.org/TR/xquery/#id-conformance
[4] http://www.w3.org/TR/xquery/#N401EFC
|
