> David Morris writes:
>
> > I missed the restriction of figures to logical graphical objects (text
> > only as an alternative for non-graphical rendering). I guess this
> > would be OK if we are clear that a table can consist of 1 row and
> > column.
>
> duhh! Please look at the <table> element for how to represent tables.
> Note that both tables and figures can have captions.
Thought I had ... still don't see any hint that a captioned box
would be considered a table. I really don't understand why there
needs to be any distinction between tables and figures beyond a desire
to produce diffent meta lists in some rendering of the document or
use different style details. Figures appear in figure lists and
tables appear in table lists. Tables need more support with things
like row/column labeling. But a single cell table w/o headings
should be identical to a figure. What we have then is a presentation
area in each cell which can contain most any other markup.
Decouple the graphical image from the figure and instead either extend
<img> or introduce a new tag pair like <graphic src=...> </graphic>
where the content of the graphic would be most any other html markup
except another graphic. Either the image or the text would be presented.
The <graphic> tags could enclose additional tags to define details of
graphical hot spot recognition as well as serving as anchors for the
text rendering if desired. One figure/table could contain one or
more graphic and I see no reason why graphic tags need be restricted
to table/figures.
> > While on this subject ... figures should have optional borders and
> > tables need to have optional cell divider lines (cell borders?). For
> > cell borders, it might work to have attributes on the <tr> and cell
> > tags (I'm too much of a novice reading DTDs, can't figure out what
> > the cell tag is).
>
> So far I haven't had any requests for borders on figures, so we could
> add this if there is enough demand. As for tables, I have kept to the
> crude borders/no borders and want to leave the details to linked style
> sheets. It will be possible to add attributes to control the border
> style directly, but this is a slippery slope and its not obvious how
> far to proceed down it.
I have perhaps missed seeing any proposals on style sheets beyond the
various mentions of the generic idea. I believe that a style sheet
should simply specify defaults for attributes which can be specified
locally on individual tags. Then the style sheet specification only need
deal with mechanics of representation of the style information, merging
of the various attributes from the hierarchy of sources, etc. The
effect of all options will be carefully described in the context of
the effected tags. It would be frustrating to have to create
separate style sheets for ocassional localized deviations. I'll
leave the remainder of my thoughts til style sheets are under serious
discussion.