[
Lists Home |
Date Index |
Thread Index
]
I prefer out-of-band mechanisms for this sort of thing. For one, why presume
that XML Schema types are the only metadata that will be useful for an
application? The practice of annotating the instance itself is limiting, and
inhibits generalized processing of information content (since a generalized
processor cannot distinguish between the information content and annotations
on it without explicit knowledge of the types of annotations allowed).
Microsoft's XML SDK has decent APIs for exposing the XML Schema type of
nodes. I don't know why more tools don't do so. That seems to me a weakness
of the tools. But there are tools out there that let you get at that info.
Beyond that, I'm in favor of generalized approaches such as Schematron,
Schematron-like languages, and the Schema Adjunct Framework as foundations
for attaching metadata to nodes to facilitate processing. That seems to me
to afford a great deal of extensibility to support varying use cases,
without forcing any particular use cases to change to accomodate other
unrelated use cases (which any infrastructural feature runs the risk of
doing).
Annotating instances with the necessary metadata seems to me to be a
slippery slope, and can introduce tighter couplings between processing
applications (since they now have to deal with the annotations as well as
the information content).
> -----Original Message-----
> From: Mike Champion [mailto:mc@xegesis.org]
> Sent: Tuesday, March 05, 2002 1:45 PM
> To: xml-dev@lists.xml.org
> Subject: RE: [xml-dev] Stupid Question (was RE: [xml-dev] XML doesn't
> deserve its "X".)
>
>
> 3/5/2002 4:31:40 PM, "Dare Obasanjo" <dareo@microsoft.com> wrote:
>
> >What exactly was your original concern and how is it not solved by
> >xsi:type?
>
> I was basically wondering why xsi:type isn't more widely used,
> in the context of Nicolas Lehunen's lament that the PSVI isn't
> widely supported. Putting the type information in the instance
> would, in my "stupid" thinking, hit the 80:20 point by giving
> applications access to the type of an element without all sorts
> of currently unsupported voodoo. As with most things, it
> comes down to best practices -- you CAN do lots of good things
> in XML if you carefully choose bits and pieces from the specs
> and roll your own code rather than expecting the parser/validator/
> wizard/whatever to do them for you.
>
> Or is my basic assumption stupid, and people DO frequently
> use xsi:type
> rather than out-of-band schemas for this sort of thing?
|
