xml-dev - Re: Requirements for making DTD validation work with namespaces

["Paul W. Abrahams" <abrahams@valinet.com>]

Wayne Steele wrote:

> IMHO, that solution (like just about every other one I've heard) is lacking.
>
> I believe Microsoft tried something like that, with unpleasant and
> incompatible results (Hint: try to declare an element with a colon in the
> name, and no attributes, using MSXML 1.0).

My proposal did not alter element syntax at all.

> WRT the earlier proposal by Paul Abrahams, let me draw your attention to the
> following line from the XML Namespaces REC:
>    "No entity names, PI targets, or notation names contain any colons."

It's clear that any method of allowing DTD validation in conjunction with
namespaces that works with published DTDs will require some modification to the
namespace spec.  The question is whether the modifications have a negative
impact on existing documents and processors.  The rule you cite would, to be
sure, need to be changed, but that would only have the effect of making
documents namespace-valid that are not currently namespace-valid.

> Everyone seems to have their own 'pet' schemes for making DTD validation
> work with Namespaces.
> To help further useful discussion of this, I propose the following:
>
> ====================
> Requirements for making DTD validation work with namespaces
> ====================
>
> I submit that any "special" processor behavior needs to be 100% compatible
> with XML 1.0 and the XML Namespaces Rec:
>    1. All XML 1.0 Valid documents are still Valid;
>    2. All XML 1.0 Invalid documents are still Invalid;

I don't see why that's an important requirement.  But the question of whether an
XML document is valid depends on what it means to retrieve an entity.  I'm
assuming additional processing at that stage, but that's perfectly consistent
with XML 1.0.

>    3. All namespace declarations work just as in the XML Namespaces REC
> (whether document is Valid or not)
>
>    A frequent proposal is creation of a new definition of "Namespace-Valid"
> XML Documents.
> I also submit that, for this to work:
>
>    4. ALL XML 1.0 documents that are Valid and conform to the XML Namespaces
> REC, are considered to be "Namespace-Valid";

My proposal has no difficulty with that, assuming that the document is
XML-validated as I propose.  And if it i

>    5. SOME XML 1.0 documents that are Well-Formed, Invalid, and conform to
> the XML Namespaces REC, are considered to be "Namespace-Valid".

Since "some" can mean anything between "all" and "none", I don't see what force
this requirement has.


> I, for one, am not interested in any scheme for integrating DTDs and
> Namespaces unless it adheres to all five requirements, above.

How does my proposal fail to meet your requirements?  Those requirements don't
preclude relaxing the constraint on entity names.

> Sadly, the leading candidates seem to be:
>    1. Ignore the whole problem, and hope it goes away

It probably won't, judging by the activity in the threads on this subject.

>    2. Nested Parameter Entities, as used in the DTD for XML Schemas

I assume you're talking here about the DTD for schemas, not schemas themselves
(which are clearly in competition with DTDs).   But as you admit, nested
parameter entities are not a very good solution.

Paul Abrahams