[
Lists Home |
Date Index |
Thread Index
]
Simon St. Laurent:
> > Total passionate disagreement.
>
> Cool.
Yeah, my preferred state. :-)
> That's the problem. I'm talking about representing information, which
> is all that XML itself does. As text, at that, quite explicitly.
> You're already on to processing. Processing is a great
> topic, but very
> different from representation.
How is the XML useful if you're not processing it in some way (and don't
tell me it makes great wallpaper :-)?
> Because a single canonical representation, used the way you
> propose, is
> an enormous pain in the ass to people whose representations are either
> less canonical or differently canonical.
>
> If you want to have a schema, fine. Consider it metadata, and keep it
> out of the XML stage. Apply it to the information you get out of the
> XML, and go from there. Not really a problem.
Once again, is this a criticism of the whole notion of associating an XML
document with a schema, or with XSD itself? The other day a friend of mine
(experienced in XML document engineering) told me that "XSD has set back the
cause of XML schema by five years." I'm starting to fear that this is the
case. I am certainly not someone who is going to stand up to defend XSD.
It's really terrible (despite the fact that the people who worked on it are
all very intelligent, experienced and hardworking). At the same time, we
*need* schemas if XML is to achieve anything near its potential.
What we do with XML is to automatically generate user interfaces, program
code and database bindings based on XML schemas that encapsulate business
semantics. We extend the schemas with additional metadata when necessary,
but we also rely heavily on the basic information in the schema to make this
possible. The result is a rich user experience, since we can do client-side
type checking and the like. Incidentally we use SOX, not XSD, and while it
is infinitely preferable in most ways, the lack of a floating point type,
for example, is sorely felt.
What are you doing and why don't you need a schema (or more to the point,
why is the schema actively harmful)?
> Sure. Typing is very useful for efficient information processing when
> computers are not so good at multiplying strings. It's a straitjacket
> for information representation when processing constraints don't apply
> to the representation itself.
>
> XSD's integration with other XML specs is certainly worth
> criticism, but
> strong typing of XML itself seems, well, just inappropriate and maybe
> even silly.
I haven't even read the XPath 2.0 spec yet. It looked pretty scary coming
out of the printer just now. But the tight integration of XSD with other
specs, which frightens me, shouldn't be confused with the overall potential
of XML schema to enable a whole new class of XML processing apps.
Matt
|
