xml-dev - Re: Options in XML 1.0

["Simon St.Laurent" <simonstl@simonstl.com>]

At 12:56 PM 11/10/00 -0500, Jonathan Borden wrote:
>Simon St.Laurent wrote:
>> I'm not denying that these things are useful in certain cases, but I'm
>> questioning whether the way they are implemented in XML 1.0 is sensible
>> and/or safe.  I have no problem with 'pre-validation transformation',
>> provided that it takes place in a clearly demarcated processing space and
>> isn't jumbled with basic XML document reading.
>>
>
>Fortunately, or unfortunately depending on your perspective, XML 1.0 is
>already specified and has been so for some time now. Like it or not DTDs are
>a part of XML 1.0. Perhaps not XML 2.0, but I am not writing standards that
>specify XML 2.0 yet, and until I can compare (the theoretical) XML 2.0
>against the current XML 1.0 specification I can't make that distinction.

Yes, XML is already specified, and has been for over two years.  I'd like
to suggest that we've learned something over those two years.

David Megginson once said that he hoped successive XML specs would get
smaller and smaller.  That doesn't seem to be the case, but one might hope.

>So like it or not, XML 1.0 specifies the procedure by which attributes are
>defaulted in an internal subset and leaves parsing of the external subset
>optional. It is true that depending on how a non-validating parser is
>implemented, the same XML 1.0 document will produce different groves or
>infosets. Many  sensible people prefer that a non-validating parser produce
>the same grove or infoset as a validating parser and I for one would not use
>a parser that did not produce the infoset I was expecting or changed the
>infoset depending on whether the validation option is turned on or off.

If you have control over which parsers are used to read your documents,
it's not a problem.  If, however, you don't, it is a problem, and not one
easily solved.

>Unfortunately for your position, part of the pre-validation transformation
>that may occur during the parse phase (i.e. PE substitutions and attribute
>defaulting) may indeed be 'jumbled with basic XML document reading' but this
>is a well known and frequently used feature of XML 1.0. 

Feature or flaw appears to be a matter of perspective.

>XML 1.0 conformant
>parsers must support this. I agree with David Megginson as it was important
>that the subsetting of SGML to form XML was associated with a distinct name
>change (XML), that any subsetting of XML also be associated with a distinct
>name change. For example CML or SML is less confusing than Common XML or
>Simple XML and as with SGML and XML there is indeed no reason not to support
>even more simple markup languages.

I wasn't discussing naming here. In fact I'm questioning the wisdom of XML
1.0 itself here, not proposing an alternative.

>What is important is that we all agree on what XML 1.0 is, and given its
>precise specification I think there is precious little room for argument
>regarding this.

'Given its precise specification' there is plenty of room for argument
about how wise that specification actually was.  I think you're confusing
different categories of interoperability problems.

Simon St.Laurent
XML Elements of Style / XML: A Primer, 2nd Ed.
XHTML: Migrating Toward XML
http://www.simonstl.com - XML essays and books