Mark Needleman - DRA wrote:
> besides bath and dc (or zxd as Rlaph wants to call it) im wondering if we
> want to define something for all of the other public attribute sets out
> there (or some subset like cimi gils, etc)?
I assume you mean for other *profiles* (we definitely don't want index sets
corresponding to attribute sets).
Before we know whether we want that I think more analysis is needed. I
proposed the Bath set just to get discussion going, but I'm not sure, even if
we agree that we want all those abstract indexes, that we want to roll them
into a single index set called Bath.
There's the question of Dublin Core. Ralph suggests using the prefix zxd.
Now that has me wondering (Ralph?), is this because you want the dc prefix
for something else (unrelated to Dublin Core) or because you want this set to
be something different that Dublin Core? (In the same sense that the XD
(cross domain) attribute set is sort of but not exactly Dublin Core.)
If it's the latter then we need some serious discussion. I think that we
need to clearly define abstract indexes for Dublin Core. If someone ask how
to search for a DC Title (as opposed to a Bath title) we need to point them
unambiguously to a specific abstract index. If we have something different
in mind for a cross domain search, then I suggest that we need different
index sets -- one DC and another XSD.
Now, to the CIMI/GILS profiles. Bath actually defines very specific attribute
combinations, which (leaving out the relation attribute) correspond directly
to the abstract indexes we're talking about in srw. And Bath gives them
names. This makes it easy to come up with a Bath set.
CIMI doesn't really do this.
There's an attribute set and discussion of using certain ones with certain
other ones, but no specific combinations (far as I can see). So I think we
would need to consult with a CIMI person, to see if there are specific
abstract indexes that should be represented in srw. Same for GILS, I
suppose. And if there are specific well-defined searches then I do think we
should represent them in one or more well-known index sets. (I think it's
premature to discuss how many or what prefixes to use.)