No one seems so far to have addressed the point made in:

> We can't figure out how to indicate the page breaks that occur
> in the endnote section - if we use <pb n="x">, as we do in the main
> text, we'll end up with page numbers in the notes that are higher than
> the page numbers in the main text where they occur

I took "indicate" to mean "display", rather than encode, i.e that this was
mainly about rendering markup, not deciding what the markup should be. If
so, there have been sensible responses about how not to display page breaks
at at all within the footnote text, and that may meet the need in this case.
But sometimes that won't do.

One case I've commonly encountered:
a) A critical edition of a text has, in the print version, variant apparatus
at the foot of each page, and commentary in lengthy footnotes in the rear
matter.
b) For convenience of on-screen viewing while retaining the seemingly
inextirpable user expectation that it should be possible to view the
digitised text in print-edition page units, the footnote text is rendered on
screen at the bottom of the page to which it belongs, below the variant
apparatus.
c) Users expect to be able to locate references to the print edition in the
on-line version.

The snag is that such references may quite commonly be to points made in the
footnotes: and authors use two quite distinct styles for so doing:  (i) fn
NN1 (to) p.NN2, where the page number cited is the one on which the footnote
reference marker, not the note itself appears (ii) fn NN4 (on) p.NN6 where
the page number cited is that where the actual note text is to be found. The
only way I know of to allow both forms of reference to be interpretable in
the on-screen version is to indicate within the footnote material both the
page breaks *and* the page numbers (from the rear matter) concerned. Now
this will inevitably result in a display of say p.25 having beneath it a
footnote that contains reference to a page-break preceding, say, p.356. This
need not be a problem if it is explained in the introductory material to the
digital edition, at a place where any puzzled reader is likely to look (the
explanation can of course "live" somewhere in the teiHeader in the XML
instance). The form of words I tend to use is something like "In the print
edition, the notes are in the back matter. In this online edition, notes are
rendered below the text of the page where they are referenced. Where a page
break occurs within such notes in the print edition, the page number given
refers to the original location of the note, and hence differs from the page
number of the body text being displayed."

Given current XML rendering and transformation technologies, there is seldom
any need to let processing dictate encoding so long as processing is indeed
confined to rendering. Indexation, searching and retrieval may, however,
raise different demands, and there indeed it may be legitimate, or even
necessary, to let such demands influence, if not wholly determine, encoding
practices. But that's not the issue here.

Michael Beddow