Stephen Deach wrote:

 > This whole debate seems to miss several key points:
 >  1.) Stripping the original semantics is NOT ALWAYS harmful:
 >      a.) Storing any derived form of original content
 >          that strips the semantic meaning MAY be harmful. If one
 >          actually needs the original semantics, it probably is harmful.

Agree.

 >      b.) Conversely, almost all books contain significant derived information.
 >          The editing and filtering (authoring) process often ADDS value,
 >          yet it clearly strips all the original semantics.

The loss of semantics during a book's editing/proof-reading process is
a bug, not a feature. I have recently completed such a process and
insisted that the publisher retained semantics in all steps. As a
result, the book will probably be available in enriched HTML before
it's on paper. (An Adobe product -- FrameMaker -- formed the backbone
of this process.)

 >  I can't unconditionally side with a) or with b), it depends on what I am
 > looking for in producing or in using the document.

So, you are saying that publishing formatting objects is the right
solution in some cases?

 >  2.) You can't determine the semantics from a DTD alone, you can only
 > derive the syntax and the allowed organization (of both elements (records,
 > fields, objects) and attributes (properties, qualifiers, modifiers,
 > constraints, values).

Correct. That's why my document says:

 "Publishing semantically rich XML should be encouraged when the
  semantics is globally known, e.g. MathML. Publishing arbitrary XML
  should be discouraged."

[1] http://www.operasoftware.com/people/howcome/1999/foch.html

 >  3.) A tagset (such as XSL's FOs) which is public, stable, and well
 > understood (with published semantics) is FAR better than a proprietary
 > encoding or a proprietary tagset (without published semantics).

Perhaps, but they're both bad when used on the Web. The right solution
is to publish documents in tagsets that are:

 - globally understood
 - semantic, rather than presentational

 >  4.) XML allows the definition of application specific tagsets. XSL's FO
 > tagset is designed solely for describing paginated and non-paginated
 > presentations, it is no more dangerous than any other tagset.

For reasons outlined in the paper, I believe XFO is much more
dangerous than other, more abstract tagsets developed within W3C.

I also think it's a significant departure from SGML's tradition of
capturing semantics, not presentation.

-h&kon

Håkon Wium Lie             http://www.operasoftware.com/people/howcome
howcome@xxxxxxxxxxxxxxxxx                      simply a better browser



 XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list