Lists Home |
Date Index |
Microsoft has been in the business of providing virtual XML views over relational stores since the turn of the decade when we shipped SQLXML 1.0. We also spent some time investigating XQuery over relational [and other] data stores.
What you call FUD, I call the voice of experience.
PITHY WORDS OF WISDOM
Eat right, Exercise, Die anyway.
From: Jonathan Robie [mailto:firstname.lastname@example.org]
Sent: Tue 12/14/2004 8:16 AM
To: Dare Obasanjo
Cc: Ronald Bourret; XML Developers List
Subject: Re: [xml-dev] What niche is XQuery targeting?
Dare Obasanjo wrote:
> People keep saying this as if it so trivial. Basically all you've
> described is what Crystal Reports does today but you've substituted
> SQL for with XQuery. However due to the fact that XQuery + XSD have a
> significantly different type system from your underlying relational
> or other data stores this causes for interesting impedence mismatch
> when performing queries.
Hey, web messages require XML in complex formats. Web pages that combine
database and other data are often created by first building an XML
structure, then transforming it to HTML. Getting this data in XML format
matters. And one reason we describe it as trivial is that some of us
have been writing queries that do this, we've been executing them
against working systems, and we're happy with the result.
> Heck, the XML Web Services folks are choking on figuring out how to
> map XSD to variables in OO languages let alone how to map an entire
> query language with operator precedence rules, collation rules, sort
> order, etc from one to another.
> Treat everything as a virtual XML document then query it with XQuery
> is one of those ideas that sounds good on paper until you actually
> start trying to implement it or use it.
FUD alert. You are right when you say that implementation is
non-trivial. Not everyone can write such an implementation. But it's not
hard to use such an implementation.
When you actually start playing with an implementation that is good for
both XML and relational sources, this idea feels very good indeed, and
you don't have to worry about the fact that the implementors had to