Lists Home |
Date Index |
At 05:02 PM 12/4/2002 -0500, Mike Champion wrote:
>On Wed, 04 Dec 2002 16:33:26 -0500, Jonathan Robie
><jonathan.robie@datadirect- technologies.com> wrote:
>>I think John Cowan's text implies that there might be problems with reuse
>>if there are datatypes in the data and XQuery is being used to process
>>it. I would be interested in seeing some concrete examples that
>>illustrate this point.
>If you start with a text mindset and add datatypes to make life easy for
>machines, it's hard to imagine a problem ... maybe Rich Salz's scenario
>if a query was looking for a floating point value and the instance was too
>long to fit in a floating point representation, I don't know enough about
>XQuery to know if it would work or not, or if it's implementation-dependent.
I'm not sure that a valid XML document can ever exhibit the problem Rich
mentions. Rich seems to say that constraints on the minimum and maximum
size of a float are not enforced by XML Schema because they are constraints
on the value space. I have not had time to check, but I believe constraints
on values in the value space like this are enforced. If Rich is right, the
solution would be to make datatypes more restrictive than Rich thinks they
are, which is not what he seems to want!
>The problem is if you start with an object mindset or an RDBMS mindset
>and think of XML as merely an implementation detail for some wizard to
>worry about. That leads to situations such as we see with Word's
>HTML format (and presumably its future XML format) -- you get stuff
>that is querable in principle, but so nasty to work with in practice
>(e.g., it requires microparsing of the element or attribute values) so as to
>make 3rd party queries impractical.
I can't see how a wizard would be very useful for this kind of data.
>Again, the issue isn't datatypes per se, but getting so wrapped up in
>the world of proprietary wizards and GUIs that you lose sight and control
>of the XML itself. My concern is that XQuery, by starting from
>the premise that it is a "strongly typed language" (not to mention one
>that is too complex for ordinary mortals to fully comprehend), will
>encourage reliance on the Wizard behind the curtain rather than honest
>proletarian labor :-)
XQuery is not a strongly typed language in the way that standard
programming languages are. Partial understanding, untyped data, and data
with various levels of typing were all fundamental to the design of XQuery.
Are there particular kinds of scenarios where you think the typing of
XQuery might get in the way? Perhaps I can try to provide some queries that
perform certain tasks where you are concerned, if you can describe what
those queries should do.
>To switch back to Uche's metaphor, the concern is that XQuery declares
>itself part of the Aristrocracy, and says of the Bohemians "let them
>eat cake" [I've been trying to work in that line all afternoon, sorry!]
>Uche is saying that if you want real interoperability across platforms
>and across tools and across time, you are better off sticking with the
>Bohemian diet of bread and avoiding the cake -- plain old text and
>markup that anyone can understand. But this discussion is basically
>pointless in the sense that there's no necessity determined by the
>technology, it's all in the mindset that the technology facilitates ...
>which can be resisted if you're careful enough.
I get lost in all the discussion of Aristocracy and Bohemians, bread and
cake. That's why I try to work with concrete scenarios, they help me
understand the problem domain better.