>I'm writing a book about TEI and SGML and I need to explain how to use
This is good news! I hope you will share some samples of the draft with
>the element for the trascrption of primary sources.
I assume you have already seen my former colleague Peter Robinson's
excellent book on the subject -- if not please let me (or him!) know.
>I've noticed that the resp attribute has been defined sometimes as
>CDATA and other times as IDREF. The Guidelines don't mention anything
>about the necessity to link it to the ID of respStmt or similar
>elements in the header. On the contary somewhare is stated that it can
>contain a literal string describing tha nature of the responsible for
>a particular encoding.
There is indeed some inconsistency in the way the resp element is
defined in some of the editorial tags. In general, if something has a
declared value of IDREF, then it *must* refer to some other element,
which will often (as you suggest) be in the header of the text. If it
has a declared value of CDATA however, there is no such guarantee.
>Having a look inside the DTD I've seen that this double definition
>is true. For instance the <sic> element has the following
>declaration in the teicore2.dtd file:
>The two elements (sic and corr) are simmetric and I can't see any reason to
>distinguish the data type of their resp attribute. Is this a bug or,
>if it is correct, what is the reason for this double form?
This however is emphatically not a bug, it is a deliberate design
decision. Some times people want to emphasize the original, and
sometimes the correction. We decided to provide for both tagging
options because there was a very vocal school of thought which felt that
SGML content was somehow more privileged or more visible than attribute
values. See further the discussion on pp 166-7 of TEI P3.
>Another question concerns the <hand> element. It has an hand attribute
>that should contain a "unique identifier, either numeric or alphanumeric,`
>thereafter in the document to refer to this scribe or handwriting
>style". Nontheless hand is declared as having a CDATA data type.
>So the hand attribute of all the transcription elements, which is
>declared as IDREF, can't refer to it, but must refer to the global
>attrbute ID that the <hand> element can bear. Anyway this conclusion
>must be inferred, since the Guidelines seem to support the idea that
>the hand attribute of the <hand> element is the target. Sorry for my
>obscure prose, but is not only my fault, since the homonimy is in
>the TEI itself.
Yes, this is an error. It is on the list of corrigible errors for the
next revision of TEI, so I fear you don't win a prize for this one --
but thanks for reporting it, and keep hunting for others!