We discussed this at the March 2006 meeting. It is addressed briefly (and
unfortunately, inadequately) in the meeting report,
http://www.loc.gov/standards/sru/march06-meeting/report.html
The issue raised was that in some circumstances it can be very expensive
for a server to calculate the number of records with any accuracy, but with
the current spec it has no choice, at least it has to return some value,
accurate or not. Any solution to this problem, other than via extension
(for example the client says, under the covers: "I don't really care about
record count", the server puts in a filler value, and the client ignores it)
would be disruptive to the current spec, version 1.1 or 1.2. Version 2.0
will offer a parameter level solution. We've been talking about this within
the OASIS group and will bring that discussion to this list when we've
worked out procedures for handling 2.0 issues (we're just getting started).
--Ray
-----Original Message-----
From: SRU (Search and Retrieve Via URL) Implementors [mailto:[log in to unmask]] On
Behalf Of Roger Wallin
Sent: Thursday, August 09, 2007 7:19 AM
To: [log in to unmask]
Subject: Re: numberOfRecords
Ok, but I don't agree with this being obvious:
"there doesn't seem to be a reason to introduce an extension that says that
you don't mind about the numberOfRecords."
If the client-program knows that the question will last fex. twice as long
if the server also counts (or tries to count) "numberOfRecords", shouldn't
the client-program be able to tell the server that it wants a really fast
answer and under these specific circumstances it is not interested in the
numberOfRecords?
This could be the situation for very complex queries or when the server
can't cache the client or when the server uses effective database-specific
methods to retrive the sub-pages or.......
But you might be right, perhaps it's a trifle....
Regards RogerW.
On Thu, 9 Aug 2007 12:39:49 +0200, Theo van Veen wrote:
>I'm not talking about keeping state. The numberOfRecords is returned in
>each response. I just want to say that when you return for example:
>
><srw:numberOfRecords>123</srw:numberOfRecords> or
><srw:numberOfRecords>approximately 123</srw:numberOfRecords> or
><srw:numberOfRecords>unknown</srw:numberOfRecords>
>
>there doesn't seem to be a reason to return a diagnostic or to
>introduce an extension that says that you don't mind about the
numberOfRecords.
>
>Theo
|