> My intent was to invite reflection on the possibility of reading
> from placement a difference of kind
While there certainly may be differences in kind, and there certainly
are differences in placement, in many cases there may not be a 1:1
correspondence between the two. Besides, it's hard to imagine a con-
tent object that simultaneously is an "additional comment ... marked
in some way as being out of the main textual stream" and in some
way a structural division of that textual stream.
> Would the use of the "corresp" be an alternative and appropriate
> way linking <note> elements in some of the cases identified ...
Yes. All anchored notes, whether footnotes, marginal notes, or
endnotes, are encoded this way at the WWP. (This avoids some, but not
all, of the horrific problems you get into trying to properly encode
page breaks along with catchwords, page numbers, etc., when footnotes
span page boundaries.)
Of the 3046 <anchor> elements in our text-base, only 25 do not have
corresp= attributes. Wouldn't surprise me if they are errors. Of the
4364 <note> elements, 4087 have both target= and id=, indicating that
most likely there is a corresponding element (<anchor> only if no
other element, e.g. <persName>, is clearly appropriate) that points
to it via corresp=.
> Any one have any experience processing documents marked up this
> way? Compared to those using <ptr>?
We have experience using our documents so encoded, but have never used
<ptr> for this. I am betting Lou meant "<anchor>", since a <ptr> is no
more allowed between two <l>s than is a <note>, but an <anchor> is.
 TEI P4:2002-03 section 6.8, p. 160;
 In case anyone is interested in the complete breakdown:
10 <note> with corresp= & id=
29 <note> with target=
66 <note> with id=
4087 <note> with id= & target=
In order to make this more useful I should of excluded those
<note>s that were descendants of <teiHeader>. Oh well.
P.S. I will add that the WWP has long thought that <anchor> should,
rather than sitting around looking pretty saying "you may now point to
me", have a specialized attribute (structurally analogous to targets=)
that allows it to do a little work and point back to whatever element
is pointing to it. (Thus allowing bidirectional ID/IDREF links without
using the general-purpose and thus semantically impoverished
corresp=.) Note that this desire helps satisfy the concerns of
inline-only folks; the <link> element performs this task perfectly
well as is.