I have a radical proposal to make concerning the "double-endpoint
attachment" method described in section 19.2.2 of TEI P4. I suggest
that you deprecate it and eventually withdraw it. My objections to it
are as follows:
(a) I don't see why in an encoding standard you need two different ways
of expressing textual variants when one proper way would do.
(b) It's not really an attachment method since in attaching the <app>
structure to the baseline it specifies an additional baseline variant.
"Parallel segmentation" is not an "attachment" method either since it
must be embedded inline, so there is something wrong with the
(c) It does not seem possible to have a variant spanning two variants,
or spanning one variant and the baseline.
(d) Each variant requires two document-wide unique identifiers. In
practice this would mean thousands of identifiers per document. Not
only would their uniqueness have to be verified each time the document
is loaded but also the references to them in the possibly external
apparatus would have to be checked to see if they all matched
identifiers in the base text. Even worse, any subsequent deletion of
text from the base text might create dangling references in the
(3) The use of two separate endpoints is computationally unclean.
Consider the situation where one of the endpoints is inside an element
and the other outside it:
<anchor id="a1"/>some roman text <hi rend="italic">and some italic
text<anchor id="a2"/> and more italics</hi>
<app from="a1" to="a2"><rdg>a variant version</rdg></app>
If the two anchors are used to mark the span of a variant, this can
create an overlapping structure, and since readings stand as variants
to one another this means that the baseline text becomes effectively:
<rdg>some roman text <hi rend="italic">and some italic text</rdg> and
which is neither XML nor parsable.
To these objections should be added those of the guidelines themselves:
"Because creation and interpretation of double end-point attachment
apparatus will be lengthy and difficult it is likely that they will
usually be created and examined by scholars only with mechanical
I have not heard that any mechanical assistance is in fact available.
Is anyone out there actually using "double-endpoint attachment"? I
would like to know.
I would prefer modification of the "parallel segmentation" method so
that it can handle overlapping variants.
Dr Desmond Schmidt