I sense some convergence of threads. Would the proposal of extending the tabular provisions fit in with
dicussions about floating divs? I ask because the relationship between unnumbered divs can be capture in a 
table content model. If one is mindful that a table is a mechanism for combinatorial reading (verticals, 
horizontals and even diagonals) based on a grid system for ascribing location to any particular cell. Looks 
like feature structures come home to roost:  (x, y), (x, y, z), (x, y ,z, ... n) ...

> In the general case of text segments that, for semantic as well as
> presentational reasons, need to be rendered with a specific set of
> horizontal and vertical aligments, the concept that comes to mind is
> "table". The snag is that the authors of what is now in P5  called the
> Performance Texts chapter decided that cast lists were a special case of a
> TEI list, not a TEI table. Which means the existing tagset for encoding a
> cast list doesn't offer a crucial facility that tabular markup brings,
> namely the ability to give strong hints about vertical and horizontal
> arrangement and alignment to a processor, without having to make the
> alieneation tail wag the rest of the poor dog into a dizzying blur. Time to
> extend, I'd say.
> Michael Beddow

Francois Lachance, Scholar-at-large

Everyone is a little bit crazy; everyone at some time has a learning disability;
No one is ever a little bit positive.