W3C home > Mailing lists > Public > www-zig@w3.org > December 2002

Re: Version 1.0 of SRW and CQL

From: Archie Warnock <warnock@awcubed.com>
Date: Mon, 2 Dec 2002 11:52:25 -0500
Message-ID: <02f501c29a23$317a59b0$0901a8c0@pommard>
To: "Mike Taylor" <mike@seatbooker.net>
Cc: <www-zig@w3.org>

----- Original Message -----
From: "Mike Taylor" <mike@seatbooker.net>
> The CQL syntax is designed such that you can add other relation
> keywords in without invalidating or changing the semantics of any
> previously valid search.  So it would be a simple matter to support
> (for example) ``gils.location within "10,10,20,20"'' and similar

So something like ``geo.time_period_of_content before 1998'' wouldn't break
anything - it would just be domain-specific, and not actually enumerated
within the initial specification?  Perhaps this addresses some of Eliot's
concerns?

> queries.  If there's interest, I could even tweak my CQL suite
> (http://zing.z3950.org/cql/java/) to configure its set of recognised
> relations at run-time from a configuration file.

I can see where that might be useful ;-)

> In CQL, the equivalent information to structure attributes is often
> implied by the relation or a relation modifier.  For example, the
> "exact" relation, which specifies an exact match between the supplied
> search-term and database field as opposed to a keyword search, implies
> the BIB-1 attribute 4=108 (structure=string) as opposed to phrase,
> word-list or similar.  It may be that new relations and/or relation
> modifiers can meet your need in this area.

Actually, I kind of like this approach for at least one instance we've run
into.  We've had a bit of ambiguity with the temporal relations, as some
make sense when the term is a single date and not when the term is a date
interval, and vice versa.  Tying the structure to the relation in this way
could resolve some of that ambiguity.

> > It would sure be handy for assisting the migration process to CQL.
>
> The Index Data guys and I have been toying with the idea of a "Type-1
> query" qualifier-set for CQL, in which you'd be able to specify
> qualified queries like this:
>
> "type1.1=1016,5=108"=fruit
>
> I'm not prepared yet to say anything about whether this would be A
> Good Thing or A Bad Thing (:-) but it is at least A Possible Thing.

It's probably A Good Enough Thing for right now.

Archie

-- Archie Warnock                   warnock@awcubed.com
-- A/WWW Enterprises                http://www.awcubed.com
--    As a matter of fact, I _do_ speak for my employer.
Received on Monday, 2 December 2002 11:52:31 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 29 October 2009 06:12:23 GMT