I disagree with Ray's claim about case #2. But, I do consider putting the
resultSetID as the only entry in the query to be a hack to eliminate a
parameter. If anyone else thought that just adding a resultSetID to the
request was a good idea, then I think we should seriously consider it.
> -----Original Message-----
> From: Ray Denenberg [mailto:[log in to unmask]]
> Sent: Thursday, June 20, 2002 10:36 AM
> To: [log in to unmask]
> Subject: Re: TTL
> Mike Taylor wrote:
> > If the client wants the server to re-execute the search, then it
> > should re-send the query. Referring to an existing result
> set _must_
> > mean that the client want to ... refer to the existing result set!
> I'm uneasy about distinguishing the special case where the
> query consists
> solely of the resultSetId to mean "don't execute this, it's a present
> request". If that's what we want it to mean the protocol
> should say so. Or
> (re-)consider an explicit result set parameter. There's not
> much difference
> (1) RS1
> (2) RS1 AND "cat"
> but if you send (2), and in particular, if you specify zero
> records, the
> server will likely execute the query, even re-execute it (if
> it has already
> recently executed that query). So it doesn't seem so
> outrageous an idea
> that a server might interpret a request with zero records
> specified to mean
> execute the query.