[docbook-tc] The CALS + HTML table model

[Paul Grosso <pgrosso@arbortext.com>]

At today's telcon, I took the ACTION item to summarize
the "ambiguity points" of the "union CALS+HTML" table
DTD module I sent to the list earlier [1].

HTML doesn't have a tgroup element whereas CALS requires it
as the child of the table element.  So you can tell which
table model you have by seeing if the <table> element has
a <tgroup> child or not.

The content model for <table> is basically (ignoring titles
and indexterms and such):

  (tgroup+ | 
      (caption, (col*|colgroup*), thead?, tfoot?, (tbody+|tr+)))

So there is no possible ambiguity at that point.

Both models have <thead>, <tfoot>, and <tbody>.  In the HTML case,
the content model for each is (tr+) and in the CALS case, the content
model for each is basically (row+).  So the content model in the union
DTD module for all three is basically:

  (tr+ | row+)

So that is one point of ambiguity where someone could mistakenly have, 
say, a thead in a CALS table containing tr+ instead of row+.  I believe 
this is really the ONLY point of potential content model mixing.

As far as attributes, there are the following situations:

html:table has the following attributes not recognized by the CALS model:
  summary border rules cellspacing cellpadding bgcolor width
cals:table has the following attributes not recognized by the HTML model:
  tabstyle tocentry shortentry pgwide orient colsep rowsep 

Both html:table and cals:table share the frame attribute but do not
share any values (the HTML DTD says they did this on purpose!):
  html: "void|above|below|hsides|lhs|rhs|vsides|box|border"
  cals: "top|bottom|topbot|all|sides|none"

For all of thead, tfoot, tbody, HTML has an "align" attribute
that would be ignored by the CALS model.  Furthermore, in both
models, they all have a valign attribute and in both models, the
values "top|middle|bottom" are allowed, but in the HTML model,
an additional value of "baseline" is allowed that would be
ignored by the CALS model.  

I believe that is the full set of issues where there is any possible
confusion between the two models that the proposed DTD module could
not catch.

I still feel we would be doing DocBook users a service to allow them
to have both CALS and HTML tables in a document, and I do not feel
the above issues--which we would prohibit via the documention but 
could not prohibit via the DTD--are so problematic as to cause us to
forbid the use of HTML tables.

paul

[1] http://lists.oasis-open.org/archives/docbook-tc/200204/msg00002.html