Sorry for the late response,

On 18 Jun 02, at 10:40, Alan Kent wrote:

> On Mon, Jun 17, 2002 at 10:41:27AM +0200, Theo van Veen wrote:
 ...
> ...
> > With respect to having meaningful names for resultsets (I do not
> > remember who mentioned it): the client can save previous queries and
> > link them to server-supplied resultsetids (which are hidden for the
> > user). For what reason should resultsets have menaingfull names?
>
> One clarification please to make sure we are on the same wave length.
> I can see two reasons for supporting result sets. Are both wanted?
>
> (1) To be able to fetch a record from a previous query (this is why
> you would specify a resultsetid above instead of a CQL query)
>
> (2) To be able to refine previous search results in a followup query.
> Eg: First query: title="Session ids" (given set name "1")
> Followup query: set="1" AND author="Theo"
>
> I think (1) is clearly a goal. Is (2) a goal too though? *If* it is a
> goal, then I can see benefits if result sets have nice names. The CQL
> query has to contain the set name. If people do not want to be able to
> refine queries in this way (for example, you can do it by substituting
> the previous query text in instead: title="Session ids" AND
> author="Theo"), then I agree result set names can be anything the
> server feels like generating.
>
> Alan

(2) is not a goal for me. In one of our implementations we do the
refinement by showing the previous queries to the user and allow
him to add terms to a query and then we create a new query. We
also use the original query to recall previous resultsets: the users
sees the query but the software then uses the corresponding
resultsetid to request it. The user will never see or type the
resultsetid's. Only the developers do.

Theo