Eliot -- A couple observations.
1. I would like to see your document cast as a profile rather than a context
set, and the context set information extracted into a separate, more concise
document focusing on the context set definition. (I realize that existing
documents we point to as context set definitions look more like profiles,
and I'm going to make the same suggestion for those.)
The context set document wouldn't have any required/optional information,
just list the indexes (etc.) that are defined by the gils context set. A
context set definition shouldn't prescribe that a value is mandatory or
optional. That's for a profile. It also shouldn't list any of the other
indexes (etc.) -- cql, rec, dc -- not in the definition tables anyway. (It's
fine to explain how the gils values might be used in conjunction with dc or
cql values as examples or in an explanatory column, but it appears at first
glance that these values are being defined by this document, which they're
not.)
2. In view of this effort and recent related discussions (including our
brief phone conversation a few days ago, Rob's comments, the discussion of
metadata vs. data) I hope we will continue thinking about how we might
qualify indexes in cql. Rob suggested an indicator of "metadataness"; I
think we should develop something more general that would serve also as a
combination functional/semantic qualifier, possibly allowing us to avoid
defining multiple similar indexes, as for example gils distributorCountry
and contactCountry.
I don't want to open up a Z39.50/attribute architecture bag of worms, nor am
I suggesting that those aren't separate indexes in real life. But we did go
to some length in the attribute architecture work to address this and I
wonder if it would be worthwhile to think about a solution based in part on
that work.
Personally I think we spent an inordinate amount of time and energy (in the
Z39.50 attribute architecture work), while crafting the semantic qualifier,
trying to articulate the distinction between a semantic and functional
qualifier. That's part of the reason it never caught on, in my view.
But if "county" is common to many applications, and "distributor" is more or
less specific to GILS, doesn't it seem like we should be able to qualify
"country" by "distributor", with "country" defined in a general context and
"distributor" defined by the GILS context? I don't know what to call such a
modifier (semantic, functional,...) and I'm not real comfortable with
casting it as a relation modifier (it really should be an "index modifier"
but we don't have those, so perhaps such a thing could wait for the next
version, and meawhile for 1.1 use relation modifier). Couldn't
dc.description be modified by gils.resource to refer to gils
resourceDescription? Could date be qualified by "beginning" or "ending" and
further qualified by gils.resource?
Rob also alluded to the possibility of a cross domain (or utility) set. I
think we once discussed defining a utility context set, but I don't remember
what came of that discussion. I think we should pursue it.
Don't take any of this as concrete suggestions, just something to think
about.
--Ray
----- Original Message -----
From: "Eliot Christian" <[log in to unmask]>
To: <[log in to unmask]>
Sent: Saturday, August 14, 2004 6:26 PM
Subject: Context Set for GILS
> My attempt to make a context set document for the treatment
> of GILS in CQL is at http://www.gils.net/context-set.html .
>
> Cases of multiple index names with the same semantics
> (e.g., author, creator, originator), are shown in the same
> row to have these treated as exact aliases.
>
> Also, you may notice a couple of new elements that are emerging
> from the Categorization of Government Information work under
> the E-Government Act.
>
> Please do let me know whatever I have gotten wrong!
>
> (Please respond to my normal e-mail [log in to unmask] or
> to [log in to unmask] )
>
> Thanks,
>
> Eliot
|