It was always my understanding that the result set id should be supplied by
the server and that it was optional.  If no result set id, then the client
has to resubmit the search.  I don't think that that is complicated.  We
should always ask - could we convince Google and Amazon to implement the
server if we expected xxx?  I'm sure they would rather supply their name
than be expected to echo one supplied by the client.  However, I'm prepared
to live with the server always supplying a name.  But it's the server that
supplies it.  The client can rename it for the user interface.


-----Original Message-----
From: Ray Denenberg [mailto:[log in to unmask]]
Sent: Monday, 17 June 2002 16:11
To: [log in to unmask]
Subject: Re: result set model for srw

Ralph --
(1) what does "immutable" mean?
(2) do you still feel that the result set id should be assigned by the
(3) is the ttl just a "hint" or is it stronger than that?  I infered from
one of
Matthew's messages that he thinks that a server is morally bound to keep the
result set available for at least approximately the ttl period (and of
morally bound doesn't mean legally bound).


"LeVan,Ralph" wrote:

> I have no problem with saying that a result set name must always be
> and must be immutable (that's a Java word!)  It doesn't need to be
> persistent and the server will give a hint as to how long it will remain.
> think we must be compliant with the current z39.50 model which says that a
> result set is a surrogate for an ordered set of records.
> Ralph