> The aim of our group was indeed to model bibliographic items (books,
 > articles, series, journals, etc.) and their (potentially recursive)
 > relationships (e.g. 'is contained in', 'is a reprint of', 'is a
 > review of') in a uniform way.

I think the Guidelines were very much in need of a proposal along these
lines. In Utrecht, using the corresp-attribute, we have something of a
do-it-yourself bibliographic database, and a standard way to encode
bibliographic relations is clearly very useful.

Some remarks:

(1) I'm not sure about the need for the <relatedBiblItem>. Can't we just
use nested <biblItem>'s? A <biblItem> within a <biblItem> will
necessarily be a related <biblItem>.

(2) If you decide to point to, rather than include, the description of
the related item, why not use the available <ref> element? It has a
type-attribute which can carry values like 'host', 'original', etc.

(3) I would not be very happy with a proposal to remove <biblStruct>.
For us, it would create an unanticipated conversion effort in moving to
P5 (not perhaps in the TEI files themselves, but certainly in the