Closed ttasovac closed 6 years ago
I used to be against this idea — after all that's what we have inheritance for. But I'm now i'm in favor of it because I'm thinking of the use cases in which we pool different dictionaries together... it's just easier to filter entries based on their language directly, then by going up and down the hierarchy chain.
I know it's also not hard to do /TEI[@xml:lang='de']//entry
but I think we can err on the side of overexplicitness in this case.
But TEI Lex-0 is a baseline encoding for dictionaries, not just for entries. For a single dictionary, @xml:lang
on //text
should be required (that's probably uncontroversial because it's the minimum requirement to make inheritance work in the first place). @xml:lang
being optional on //entry
would still allow for the anticipated pooling use case because you can always make the inheritance explicit in your sources by simply factoring out //text/@xml:lang
to every (possibly embedded) entry
. This would still result in a TEI Lex-0 conforming mark-up. We just shouldn't force everyone to cater for this use case.
Following your rationale we would have to force the application of @xml:lang
on many other elements as well. People could be tempted to pool etymologies or //form[@type="headwords"]
(which is actually a rather common use case, too) and many more things within //entry
.
The argument was to make explicit the object language of the corresponding entry, rather than letting the ambiguous (in that it is intended for the working language) @xml:lang be inherited "by accident". Doing so, we have entries which have some global autonomy and well documented from a lexicographic point of view.
I completely understand @xlhrld's reservations, but I think it's also important for us to think beyond the dictionary as an XML document. This is what "global autonomy" of the entry means to me, whether we're just quoting one entry, or converting the dictionary to a relational database (god forbid! 😄) ...
I'm fine with this »global autonomy of entries« – as long as this is prominently stated as a major objective of TEI Lex-0 in the spec. Otherwise it just smells like an arbitrary privilege for entry
.
This would make sense when there were articles in different languages comprised in one resource which doesn't happen very often. I'd rather suggest to require
TEI/text/@xml:lang
while still optionally allowingxml:lang
onentry
. We rely on inheritance all over the place (e.g. see the examples withcit[@type="etymon"]
) so why not withxml:lang
here?