> It seems to me that a perfectly simple and well-defined interpretation
> of ems in table column widths is that they refer to ems at the start
> of the table - i.e. the type size in use when <table...> was encountered.
I agree ... proportional columns are a nice concept on the printed page
where there is a reasonable expectation that things will be readible ...
but with web browsers it will be necessary at times for the author to
dictate that this material needs to be some minimum width to be readible.
At times the material will require horizontal scrolling but the integrity
of the relationships between cells will make sense to the user. I guess
I would be satisfied to specify the minimum width in ems or some other
measure of readiblity in the user's context and then use of proportional
sizes in the table but that seems overly obtuse.
I believe there are valid arguments why a measure like ems is useful and
also why proportional sizes are good. I've heard no argument as to why
we shouldn't allow both (even within a single row!).
David Morris