> Stuart, how did this approach lead to poorly formed XHTML? There's
> nothing inherent about the TEI that should, surely? In a P4 SGML
> document I used a syntax that anticipated choice in this context to a
> certain extent and even with the rudimentary pre XSLT tools available to
> me at the time had no trouble getting valid XHTML out of it. Or is just
> a bug still needing to be squashed in your sheets?

The problem is that in TEI you can have block-level tags (i.e. div) inside sic/corr/del/ins elements inside p/span elements

HTML / XHTML do not let you nest in these ways.

The classic example is the when an editor inserts a "end-paragraph" marker within a p. Very hard to handle in XSLT.

I'm aware that these are theoretically possible, essentially by closing and reopening of the block-level tags, but we have found this approach to be fragile and error-prone. Indeed, we'll shortly be going live with updates which remove our current system of breaking block-level tags for page-breaks for exactly this reason,  instead we float the page-break into the margin. [Thanks to Jamie who's doing the actual work on this.]