Simple XML conformance

[Peter Murray-Rust]

At 03:22 PM 1/17/00 -0500, Simon St.Laurent wrote:
>At 03:21 PM 1/17/00 -0500, John Cowan wrote:
>>Peter Murray-Rust scripsit:
>>
>>> 	- tool threw a fatal error because <!DOCTYPE was missing
>>
>>The XML Rec does not forbid a parser to treat validation errors as fatal.
>>
>>> 	- REC-xml and DOM specify DTD but spec.dtd is not mounted
>>
>>Non-validating parsers that read DTDs are entitled to get upset
>>if the DTD cannot be read.
>>
>>> 	- several tools regard the absence of a DTD as a fatal error (i.e. they
>>> appear to be validating by default).
>>
>>Again, that behvaior is not forbidden.
>
>I don't think Peter's point is that these behaviors are forbidden, but
>rather that they make interoperability difficult.  Perhaps a few more
>things should have been forbidden. 

This is exactly the point! We are in slight danger of going back towards
the problems
of SGML ("you didn't add the SGML declaration????"). There are clearly
times when 
validation is required and incorrect XML instances should throw fatal
errors. The main 
problems seems to be that:
	- there are potentially five categories of parsers (from the OASIS/NIST
document which
I had read, and found extremely useful :-)
	- most parsers do not explicitly announce which category they fall into
	- some parsers additionally modify their behaviour by what is presented to
them (e.g 
		<!DOCTYPE present? external DTD subset declared?
	- there is no easy way of switching the behaviour of some parsers (i.e.
you have to 
		use a validating parser or a non-validating one explicitly.
	- some tools contain parser functionality which is unclear

I have been through this with SGML and failed to win converts. It seems not
unreasonable
most tools should be able to cope with DOCTYPE-less documents. Otherwise we
get
back to the scenario:
	"This XML file only processable with a parser of type X"

and that is very difficult to present to newcomers. 

Simon wrote:
>
>Peter wrote:
>>Henry and I are obviously keen to show that XML is simple to use with the
>>correct tools and that interoperability is achievable. 
>
>I'd love to show that as well, but lately I seem to be writing books that
>detail the possible pitfalls and suggest strategies for avoiding them
>rather than claiming XML makes interoperability easy.

I agree with this - unfortunately. As we move into the area of complex
documents
with multiple namespaces and no formal DTD, surely some of the current
range of 
parsers will have problems? Do not most XSL documents fall into this category?

>
>Simon St.Laurent
>XML Elements of Style / XML: A Primer, 2nd Ed.
>Building XML Applications
>Inside XML DTDs: Scientific and Technical
>Cookies / Sharing Bandwidth
>http://www.simonstl.com
>
>xml-dev: A list for W3C XML Developers. To post, mailto:xml-dev at ic.ac.uk
>Archived as: http://www.lists.ic.ac.uk/hypermail/xml-dev/ or CD-ROM/ISBN
981-02-3594-1
>Please note: New list subscriptions now closed in preparation for transfer
to OASIS.
> 

xml-dev: A list for W3C XML Developers. To post, mailto:xml-dev at ic.ac.uk
Archived as: http://www.lists.ic.ac.uk/hypermail/xml-dev/ or CD-ROM/ISBN 981-02-3594-1
Please note: New list subscriptions now closed in preparation for transfer to OASIS.