I think the normative "not recommended" is indeed more
approriate.
As far as I can see, Steve’s intent was to add informative text
using those words, not to replace the current wording. And “not
recommended” is not normative, whereas “should not” is normative,
though in a vague way (you can always claim conformance, with some
suitable explanation).
Nonetheless, accessibility shouldn't be pointed out
as being the only issue in using table markup for layout.
Semantics, performance and efficiency are also points that suffer
from the heavy and useless markup of tables. All this can impact
in the overall user experience (and not only screen reader
users').
Such arguments have often been presented, almost always without
giving any factual evidence. In some discussions, more concrete
arguments might pop up, but they typically turn out to relate to
specific ways of using layout tables – and quite often, in a manner
that makes them extend to other layout tools as well. For example,
amount of code does not really depend on using layout tables but on
the way you use them. You could have absolutely minimal HTML markup
for a table and do all width, height, alignment, font, border, and
other settings in CSS.
I’m afraid the principle “don’t use tables for layout” has become a
slogan and a meme that gets repeated, in slightly different forms,
over and over again, with some very abstract arguments given to
support it. It is often exaggerated so that even perfectly normal
data tables are explained as being layout tables, or the slogan is
simplified to “don’t use tables”.
The current normative text says that tables should not be used as
layout aids. It’s the first sentence there. I don’t see how much
stronger you could put it, unless you want to say “shall not”. The
text is rather abstract and it exaggerates a lot, but I don’t think
it’s realistic to open the case.
It would be a major effort to write a description of the negative
consequences of layout tables, if it would be based on facts, backed
up with references, and written in such a manner that some kind of
consensus-like acceptance could be achieved. And it would be a
matter of careful study and discussions, not just adding a few
explanatory notes. It would need to start with a definition of what
“layout table” really means – e.g., quite often even a table of
controls and labels for them is accused of being a “layout table”.
I’m pretty sure it would be difficult to reach any consensus even on
the definition.
So I think it would wasted effort to add informative text.
-- Yucca, http://www.cs.tut.fi/~jkorpela/