[Biojava-l] request/proposal for change to BlastLike DTD

[Christopher Lanczycki]

Mat,
Opinions?  I can give one of those, although authoritative it's not.

I think schemas are an improvement over DTD based on precision of
specification and the self-validating nature of schema as XML docs
themselves.  I haven't found them to be so hugely more difficult
to deal w/ in simple cases than DTDs, but they are bulkier text-wise.
As you've found, for something as large as the BioJava DTD, though, 
the DTD-->schema transition doesn't appear to 'scale up' smoothly.

At present, my biggest issue is w/ tools and readily available parsers
(preferably of the open-source variety).  Xerces does have parser (DOM
& SAX) support for schema in at least v. 1.3, although it's full level
of compliance to the recommendation I'm not sure about.  Other tools -
current situation seems spotty but likely to improve quickly.

As one of the stated goals here is to not break existing users code,
however,
I wonder how one can take advantage of the features of XML (particularly
with respect to cardinality specs) that simply aren't expressable in
DTDs.  If a schema can be devised that has the existing DTD tags as
a (deprecated?) subset, say by using restriction, I think it would be a 
beneficial addition.

Chris



"Wiepert, Mathieu" wrote:
> 
> Simon,
>         Awhile ago you had mentioned changing from the DTD to a schema.
> Have you put any work into that?  Schema is now a w3C recommendation, so we
> should probably move over to that soon.  I did try a simple conversion from
> the DTD to schema with XML Spy, but that did not work, and I didn't pursue
> it.
> 
> Any other opinions on this?
> 
> -Mat
>