OASIS Mailing List ArchivesView the OASIS mailing list archive below
or browse/search using MarkMail.


Help: OASIS Mailing Lists Help | MarkMail Help



   Re: [xml-dev] FW: [ANN]: XQuery: A Guided Tour

[ Lists Home | Date Index | Thread Index ]

Dare Obasanjo wrote:

>As for XQuery being able to support other XML schema languages in future, this is nice but I'm curious as to what exactly this means in practical terms. 
It seems the XQuery does not use (declarations for)
    - facets
    - key/keyref
    - nillable
    - details of user-defined derivation hierarchies (i.e., it is 
interested in supertypes or
    subtypes only as an unordered list, not as an ordered list or 
anything: like substitution groups,
   only the name is important)

and creates its own type hierarchy for lists and unions which is 
different from WXS.
(An improvement, I presume it could be folded into WXS.)

Given all that, XQuery looks pretty good in this department. I could 
easily imagine
(unambiguous) RELAX NG (with WXS datatypes and a minimal PSVI interface) 
fitting in. 
It might be pretty difficult to have an XQuery implementation that 
supports pluggable schema
languages, of course; and the constant revalidation probably rules out a 
schema language that
may want to access the whole document (Schematron, WXS key/integrity 
constraints, etc.)

The main problem I see for XQuery is that XQueryX completely sucks.  For 
those who have
missed it, XQuery defines a nice whitespace-tokenized language, but it 
also has another syntax
that looks like XML but is not XML: different namespace rules, <![CDATA[ 
is a function that
returns a  CDATA section, <!-- a comment is a function returning a 
comment-->, no encoding
considerations.  A completely retrograde step, from what I can make out, 
and one I hope that
needlessly clouds what otherwise is looking pretty good now. XQueryX 
could be trimmed
without loss of power: scraps to the dogs under the table.

>The working group already has seen the problems caused when one group builds a "type system" (and I use the term loosely) which they believe can work for some future language only for their assumptions to come out wrong. I originally was of the mind that this was a laudable goal of XQuery and still think it is a laudable goal to have it in the data model but wonder how feasible it is to actually create an XQuery implementation with what is basically a "pluggable" type system. A validation language with pluggable datatypes (which are basically custom checking on string values) like RELAX-NG has is fairly straightforward enough but a creating a pluggable type system where you have to deal with issues like type promotion & type substitutability is a bit harder. 
Sure, but the only (public) full-featured alternative datatype system so 
far is J. Tennison's, which ultimately relies on WXS: it performs 
translations.  So the choices need not be between WXS and
non-WXS, but WXS and a more extensible form of WXS with which you can 
define your own
types.  Jenni's datatyping system really fits at the same level as 
derivation-by-list and
derivation-by-union: type promotion and substitutability of primitives 
occurs based on primitive
types' properties and facets, which Jenni's datatypes fit on top of.



News | XML in Industry | Calendar | XML Registry
Marketplace | Resources | MyXML.org | Sponsors | Privacy Statement

Copyright 2001 XML.org. This site is hosted by OASIS