Re: Information components (DTD fragments)

I apologize for bringing things up out of order but it isn't yet clear
to me what we are supposed to be discussing or the correct time to
discuss "semantic namespaces" or "lexicons". In my opinion the correct
time to discuss semantic namespaces would have been last year, not
sometime after XML 1.0 ships. Of course the ERB has the final word on
what goes into XML 1.0 so I'll try one last time to urge that this DOES
belong in XML 1.0 and then stop tilting at windmills.

Jon Bosak wrote:
> This is an extremely important topic, and under the heading of
> "lexicons" (suggested by Tim Berners-Lee) is already down on a long
> list of things we will get to *after* we take care of the business at
> hand.  The business at hand was summarized by Tim Bray on behalf of
> the ERB in a message dated April 30.  I have reproduced that message
> below for the benefit of anyone who's forgotten it.

I have read that message twice and am not clear what the WG is supposed
to be discussing. I had presumed that we were waiting for the ERB to
outline positions and options w.r.t. the three issues so that we could
get started. Two of them have been discussed at length already and the
last is quite vague, so I presumed that there was more direction
forthcoming. In the meantime, it seemed to me better to discuss
"important topics" rather than twiddle our thumbs while we wait to hear
what our options. The later we wait in bringing up these things the
harder it is to incorporate them. I wish I had thought this through last
fall, but that's the way the human mind works.
 
> Whatever eventually comes out of this discussion, it does not belong
> to XML 1.0 but to some future version.  

I think it would be bad to create a culture of "invent your own tags" in
XML 1.0 and then try to stamp it out in XML 2.0. I consider the lack of
namespace declarations to be a a major limitation on XML document's
interoperability and in very practical terms a limitation on its
usefulness. Here are XML's two benefits, as described in PCWeek:

>    Using XML, authors can create new tags at will, even very complex
>    ones. They also can use XML DTDs to validate the structure of large
>    numbers of documents, which is important when importing the data from
>    those documents into other applications.
...
>    In addition, XML will make intelligent agents easier to design and
>    deploy. Today, agent software has to jump through hoops to recognize
>    the right data points on constantly changing Web pages. With XML,
>    relevant data points can be marked with their own tags (such as
>    <price>, for example), so they're easy to find.

Unfortunately in XML version 1.0 you can have one benefit, the ability
to make your own "tags at will," or the other benefit, the
standardization of semantics. As soon as XML becomes widely available
people will experience this huge standardization gap and invent hundreds
of incompatible ways of correcting it.

All I am asking for is a syntax for declaring that tags in the document
are equivalent to tags in a known DTD. Not for translating one DTD into
another. Not for validating against a "meta-dtd". Not for context
sensitive processing. Surely this isn't going to take up much of our two
months before July 1. I think that it requires two PIs, one to import a
DTD's tag names and another to import particular tags (which can also be
used to resolve namespace conflicts). Processing tools that don't care
about the element semantics can ignore it. We could even decide not to
specify a validation level for it. The important thing is that there be
a standardized syntax for these declarations rather than dozens of
ad-hoc mechanisms.

 Paul Prescod

Received on Monday, 5 May 1997 12:23:18 UTC