I guess I got confused by Matthew's original comment because he was
actually, quite knowingly, addressing the (probably broken, or at least
somewhat naive) requirements of inexperienced metasearch builders. There
were none or few of these present at the NISO metasearch workshop -- those
guys were all fairly seasoned, and as Ralph says, they hold absolutely no
illusions about the complexity of the task at hand.

I think perhaps some of us in the early days were guilty, naively, of
overselling Z39.50 by leaving people with the impression that you could
indeed broadcast queries and expect to get useful results. It wasn't till
later, when interoperability started to happen for real and all the
semantic differences cropped up, that we realised how wrong we were. But
that's not Z39.50's problem... nowhere in its spec does it claim (I
believe) that you can do that and expect predictability. And even though I
believe our work on CQL is also inspired by Bath (which does indeed bring
us a lot closer towards useful broadcast queries -- when it's implemented),
I don't think it's possible (or maybe even desirable) to suggest to people
that SRW solves the problem anymore than Z39.50.. only rigid,
application-specific profiles can do that.

If we *do* want to meet the requirements of the lazy (or influential)
metasearchers, then we had better make damn sure that by the time that SRW
really stands up in the spotligt, we have, at the very least, a fully
developed mapping of Bath onto SRW, so we can say to software procurers
that "if you ask for Bath/SRW you will be about as friendly to
metasearchers as you possibly can".

--Sebastian

At 17:47 10-08-2003 -0400, LeVan,Ralph wrote:
>I completely disagree.
>
>Right now, the metasearchers are scraping screens and doing all sorts of
>stuff that is unique to each site.  They understand the need to customize
>the client to the server.  They absolutely understand that they can't send
>exactly the same thing to an arbitrary list of servers and expect to get
>anything useful back.
>
>Ralph
>
>-----Original Message-----
>From: Theo van Veen [mailto:[log in to unmask]]
>Sent: Sunday, August 10, 2003 4:56 PM
>To: [log in to unmask]
>Subject: Re: Betr.: Re: xPath in searchRequest
>
>
>On 8 Aug 2003 at 22:19, Matthew J. Dovey wrote:
>
> >
> > More seriously, we are straying onto metasearch requirements territory
> > here. I suspect that one of their requirements and a reason I suspect
> > that might give for not considering SRU/SRW is that they do expect a
> > single string to work on an arbitrary list of servers!
> >
>
>I agree. And I do not see any reason for not trying to meet such
>requirements. Actually I
>would expect metasearch being one of the objectives of SRU/SRW and I realise
>the
>danger of not meeting such requirements.
>
>Theo

--
Sebastian Hammer, Index Data <http://www.indexdata.dk/>
Ph: +45 3341 0100, Fax: +45 3341 0101