Re: tab-width specifications in HTML

Peter Flynn (pflynn@curia.ucc.ie)
Thu, 9 Feb 95 10:38:42 EST

Jon writes:
Agreed that a table is (i.e., should be) "an ordered collection of
information", it is often not _just_ that; sometimes it is an ordered
collection of information plus some explicit formatting hints
[...]
either.) It's essential from the standpoint of information retrieval,
reuse, localization, etc. that tables be ordered collections of
information before they are anything else, but you have to provide
mechanisms for layering some personal, case-by-case formatting hints
on top of that foundation.

While I have a lot of sympathy with this view, it remains that HTML is
not, and never will be, a typographic specification-and-formatting
language. Hence it is unreasonable to expect it to handle the
peripheral cases Jon and I (and others) deal with in publications: if
I want serious high quality really complex tables I use TeX or 3B2 or
something.

What we need is a system to handle the vast majority of day-to-day
users' tables, which will be largely rectangular, regularly-divided,
with headings on the columns and rows, with the occasional bit of col
or row spanning. I think trying to provide more _at this stage_ is a
pointless waste of time: later on in 1995 perhaps we can consider a
more complex table model.

The only departures from this "single-value entry" kind of cell are
probably the occasions when you want either several stacked values or
some arbitrary continuous text setting in a paragraph as a cell. The
former can be handled by making the table rules a separate element
from the cells; the latter is traditionally done by predicating the
column width, or by saying "use remaining space for the text column"
or something similar. I'm as unhappy as some others about this:
suppose the user is viewing in a tall narrow window as I've seen some
of them do, or in a very shallow, wide one. I'm not sure there is a
simple answer: perhaps let the user say "take n lines over this cell,
and set the width accordingly".

///Peter