```I've never been able to see a hard and clear distinction between table as
layout and table as logical structure. At one extreme, I receive
every day texts that have been encoded using TABLEs unnecessarily--
in cases where even a little interpretation of the physical
arrangement will have revealed that the two-dimensional layout
could easily be rendered one-dimensionally (e.g. as a list) without
loss, and often with some gain. Use of TABLE in this circumstance is
irritating because lazy.

At the other extreme, there are of course 'pure,' e.g. wholly
numeric, tables.

But there are plenty of mixed tables in between. Our rough
rule is that objects should be recorded as TABLEs if the 2D layout
is significant, cannot readily be captured by other means, and/or
resists interpretation. After all, when converting books, as
opposed to authoring them, one cannot always be sure of understanding
them to the extent it would take wholly to dispense with the native
formatting and substitute another: one always needs a layout-based
fallback; and even if one DOES understand the logic of the table,
conversion to something else can be both very expensive and very,
well, disrespectful to the choices made by the author or printer,
as we find when we try to provide 1D capture of 2D features that, unlike
grid-structures (tables), have no straightforward tagging in TEI,
e.g. some uses of braces ({), mathematical formulae, etc.

As for the contents of cells, I find that though I have perhaps
grudgingly retained most of the TEI restrictions in 'legacy' dtds, in my
own, I have become much looser. In fact, I tend to define CELL and ITEM
identically, and allow both to contain P LIST TABLE L Q etc. (I also allow
TABLE LIST P and FIGURE to appear in DIVs.)

One common type of table that often contains CELLs that beg for P is
calendrical, as one might find in an almanac: the rows are days; the
columns are things like saints' feasts, roman reckonings (kalends, etc.),
phases of the moon, progression through the zodiac, and prognostications.
The last column especially, and sometimes the first, is likely to contain
discursive prose (and maybe verse too), often spreading over several days
(ROWS="3"). How pure a table is this? I do not know, but it would
certainly be inconvenient--not impossible, but inconvenient--to attempt
to capture it other than as a table.

pfs
--------------------------------------------------------------------