> > More significantly, the content model is wrong: you
> > can't have <p>s inside it. You'd have to do that with <addSpan>.
> I think (he says tentatively) that this is wrong. <ADD> does take <P>s, or
> at least it used to: "<!ELEMENT add - - ((((l | lg | p | sp), ... >"
> <ADDSPAN>, on the other hand, does not, since it is defined as EMPTY.
Ah! You are quite right... mea culpa. <add> is defined as %specialPara, so
it can indeed take <p>s (though I think your example wouldn't parse with
unmodified TEI, because of our old friend the mixed-content problem)
> > If you do want to use <add>, you could put them INSIDE the <closer> which
> > is intended to be a grouping element for all the paraphernalia at the end
> > of a <div>.
> Look again: inside the CLOSER is exactly where I put the <ADD>. (For once,
> I took the precaution of validating the samples before showing them to the
Sigh, I'm overworked. Sorry!
> > Obviously, it means the extra burden of requiring you to wrap the
> > body of the letter in a <div> as well, even if there is no postscript.
> It is not our practice to insert <DIV> levels without local
> textual warrant, even when this means that similar features are tagged
> with different levels of DIV. I.e., as we normally do things, letters with
> postscripts would get the nested <DIV2>; letters without would not. Which
> doesn't make it a less awkward 'solution.'
I'm not quite sure what "local textual warrant" means, but it does sound
as if we're converging on <div> as the least worst solution.
> > Looking at your example, I would also question the wisdom of tagging
> > distinct letters as <div1>s. A div is by definition incomplete. A complete
> > letter should be tagged as a <text>. Then you could tag the postscripts as
> > <div>s within a <back> element!
> To be sure. But as you know, I can't realistically take that route: my
> beginning point must always be our institutional practice of identifying,
> with few exceptions, the whole book as <TEXT>. IN an otherwise undivided
> book called The Collected Letters of J. Doe, each of J's letters would
> normally be, for us, a DIV1 of BODY.
Well, we've had this argument before, and clearly we'll have it again. For
my money, the collected letters of J Doe is a <group> containing a bunch
p.s. when are you guys going to join the TEI Consortium?