are you talking about the description of such a manuscript or are you
addressing the transcription? <altIdentifier> would address the former,
<div>/<ab> the latter.
Nevertheless, I think you/we need to distinguish two things:
1. The current state of "the manuscript", which will be one surface that
contains script and traces of former scripts but it will still be one
2. If we think of this object as product of some scribal activity (what
we used to describe in <origin>) the surface has two different
histories, maybe even two different sets of physical attributes as the
palimpsested pages often are reused by turning them 90° or 180°, maybe
folding them and only then writing upon the "new" surface. So, if we
consider them to be "parts" of some virtual "original" objects, we will
find that the palimpsest may have been part of another manuscript and
definitely the upper script is a manuscript of its own. Looking from
that historical perspective we will have to consider them as two objects.
I have argued before that this fact, that a palimpsest manuscript has
more than one history makes up for describing it as distinct manuscripts
using <msPart> -or, if only parts of the original, palimsested
manuscript are preserved, maybe even the newly established <msFrag>.
If you were addressing the transcription I definitely would argue for
two "texts" that relate to the same surface, e.g. using <sourceDoc> with
You may know the new "standard" IIIF  and its underlying philosophy,
SharedCanvas, offers a -I think- similar interpretation of how to think
about this question. Only right now, IIIF and TEI aren't linked together
in a defined way.
PS: In the course of a newly founded project and site concerned with
fragments  we had a meeting recently discussing this general question
as well. Stay tuned for answers how we will deal with these questions. ;-)
Am 20.06.2016 um 00:22 schrieb Marion LAMÉ:
> Dear colleagues,
> In order to follow already existing habits, we are looking for some
> examples of encoding for palimpsest texts, that is:
> - erasing intentionally and entirely or most of the surface of the TBO:
> some symbols are still observable and eventually readable on this/these
> older layer(s).
> - reusing the same surface to write another text, sometimes totally
> different from the previous one: this most recent layer on the manuscript
> is usually more readable than the older one / the olders.
> Palimpsest texts have a "semantic baggage" that is a bit different from
> similar writing processes such as the two following examples:
> - erasing part of the written surface (e.g.: a word or a letter) and
> rewrite for correcting a mistake for instance; in such an example both
> writings belong to the same layer and the writing process belongs to a same
> intention / moment / production.
> - erasing and leave this textual phenomena visible such as in the cases of
> damnatio memoriae on some inscriptions.
> We have searched in the Mark up list and TEI-L list archives and we have
> not found yet any previous messages that was convincing enough and
> definitive practice. The most interesting message seems to suggest to use
> the <altidentifier> with a @type + a value such as "palimpsest". But it
> also seems to refer to another description process, that is when one should
> collect several versions of a same text from various TBO.
> Could several <div> and / or <ab> be used to encode the overlapping texts
> on a same TBO + some attribute (@type ?) + a value ("palimpsest" ?) ?
> What are your habits describe each layer of texts that are overlapping in a
> palimpsest manuscript?
> Would some of you be kind enough to share their previous experiences (XML
> code and / or links), please?
> Marion Lamé
Torsten Schassan - Digitale Editionen, Abteilung Handschriften und
Herzog August Bibliothek, Postfach 1364, D-38299 Wolfenbuettel, Tel.:
+49-5331-808-130 (Fax -165)