#173 distinguishing two ways to encode <keywords>
The content model for <keywords> allows either for a list of keywords that apply to a particular TEI document:
(A)
<keywords>
<term>A</term>
<term>B</term>
<term>C</term>
</keywords>
(B)
<keywords>
<list>
<item>A</item>
<item>B</item>
<item>C</item>
</list>
</keyword>
However, all examples given in P5 use only (B). Since <term> seems to be defined as a word that requires definition, (A) seems inappropriate.
So I request that the Guidelines give guidance on choosing between (A) and (B). Better yet, I request that we disallow one of these so that there's a clear way to encode keywords. While we risk breaking backwards compatibility, it seems to me that the conversion between the two is trivial, and we'll save some time for anyone aggregating TEI documents from various sources.
As background, see http://listserv.brown.edu/archives/cgi-bin/wa?A2=ind1003&L=TEI-L&T=0&O=D&P=2263 .
Discussion
A first priority would be to change the definition of term to make it independant of a definition.
I don't like so much (A) since it creates an implicit list by the sole 'sibling' relation between nodes. I would also not like to lose the use of <term> which should indeed mark terms (in the sense of ISO 704). My suggestion would either to:
avide to use <term> within <item> when applicale, or,
to use <term> within <list> (but this is probably too narrow a use case to go in this direction; or we should have to consider mixing this with <gloss>, <label>, etc.)
At the NZETC we are inconsistent about how we do this (see http://www.nzetc.org/tei-source/HutColl.xml for an example).
The fact that <keywords> uses the plural, makes it an implicit list, to my eye.
I don't see what's wrong with using <term> in this way, but I agree that if we want to support it then there should be an example of its use. Of course you can also combine them: the example currently in <keywords> could equally well be organised as follows
<keywords scheme="http://classificationweb.net">
<list>
<item>Babbage, Charles</item>
<item><term>Mathematicians</term>
<term>Great Britain </term>
<term>Biography</term>
</item>
</list>
</keywords>
Lou wrote "I don't see what's wrong with using <term> this way". I believe that means he is not opposed to (A), whereas Laurent and Stuart expressed support for (B) over (A).
Unrelated to my original question, Lou suggested that the example given in the element defintion for <keywords> could be changed to to tag "Mathematicians - Great Britain - Biography" as three separate <term>s, all within one <item>. These terms comprise a compound subject heading, and determining what counts as a subject heading term (each atomic unit or the compound term) is an ontological debate that I'd like to avoid.
However, if we end up tightening up the content model of <keyword> to allow only (A) or (B), we should make sure to still allow <term> within <item> to allow for Lou's ontology of subject headings.
Council agreed to deprecate" use of <list> in <keywords>. All examples involving <keywords> in the
Guidelines will be changed to use only <term>, and prose note(s) will be added to clarify that <term> is preferred.
For future reference, note that use of <term> has the further advantage of allowing @target, with which one can point to an authority record for a term from a controlled vocabulary.
Heterodox examples duly removed. Added following comment to <keywords> reference:
<p>Each individual keyword (including compound subject headings) should
be supplied as a <gi>term</gi> element directly within the
<gi>keywords</gi> element. An alternative usage, in which each
<gi>term</gi> appears within a <gi>item</gi> inside a <gi>list</gi> is
permitted for backwards compatibility, but is deprecated.</p></remarks>