This discussion of ADQL core vs extensions reminds me of our data model
discussions. Data models, and ADQL, are both fundamental, general
technologies which can be applied in various contexts. In both cases
it is hard to say what the priorities for a given use-case are without
understanding the context, that is how the technology is being used.
I think the answer is that such core technologies should be defined in
a fairly abstract sense, without trying to prejudge how they will be
used. When the context is clear, for example when defining an actual
service such as TAP, or a distributed cross-match application, one can
say better what is core and what is an extension, what is "minimally
compliant", or "fully compliant", an so forth. Hence, in defining
ADQL we should probably not try to prejudge usage, and should rather
merely define the language, leaving issues of capabilities to be defined
when ADQL is used in some more specific context. - Doug
Received on 2006-11-02Z06:00:29