[
Lists Home |
Date Index |
Thread Index
]
As Betty said, "it depends on the tools" is one
rule of thumb. DTDs refuse to go away because
they are ultimately the bottom line of the
XML specification for basic validation. In
other words, you can usually count on them being
supported in your toolset. But given a toolset
that supports RELAX NG, I wouldn't start there
these days. Given a toolset that supports XML
Schema and a need to model, say a relational
database, I would start with XML Schema because
I can get a first cut using an ODBC-sourced
dump of the relational schema. I haven't tried
that with RELAX so I don't know how well that
works but I suspect it does.
But now I have this table-sourced XML and
I have to model relational value constraints.
At that point, I have to go to Schematron
say in the app-info elements, or I have to
move on to business logic in say C# or
Java objects.
In any of these cases, as soon as one needs
stronger datatyping, say types for integers,
one has to get beyond DTDs. Think of DTDs as
XML's bottom line bootstrapping language for
moving beyond well-formedness, and that is
a good metric, but it won't go very far
and one will likely have to if one needs
a sharable definition that side of
sharing the code. But remember, XML
only requires well-formedness to be
XML and sharing code is not out of the
question.
len
-----Original Message-----
From: tariq abdur-rahim [mailto:ecliptoid330@yahoo.com]
Interesting point! However, would the argument now,
not be that, given the realative "weakness" of DTDs in
comparison to Schemas, RELAX NG, and Schematron, what
would the purpose be of even going the DTD route for
validation? Too, given the example stated...
|
