On Fri, 4 Jun 2004, Marc Cromme wrote:
> Mike Taylor wrote:
> >>From: Marc Cromme <[log in to unmask]>

> >>2) an indication of the actual relevance value of every record in
> >>the hit set (this needs a new SOAP tag in SRW)

> >No SRW changes are needed for this: there are already two mechanisms
> >for including this information.  One is just to add it to the returned

> Less is more:
> approach a) woun't fly, since then I have to know each server
> specifically to figure out where in the returned record this information
> is buried, and in which form. What a hell of a logic to write for a
> federated search over some 30-50 targets!

More importantly, it won't fly because you can't just add the relevancy
metadata to any old schema. You'd need to create N new schemas where N is
the number of other schemas you support, just to add this one bit of
information. Bad.

> approach b) is much better, but only  if all agree that the same tag is
> used inside the extraRecordData packet to carry this information, and
> that all servers agree to encode it in some unified fashion.

Yup.  But that's the same way that anything is.


> I do not have SRW on my fingertips, but I belive that there is no such
> 'reserved' tag inside the 'extraRecordData' tag, and if it is not
> defined by the standard, we are back to the problems of approach a).

Note that my context set isn't defined by the standard either, I just
wrote it yesterday and stuck it on the web.  If people find it useful and
implement it, that's great.  If not, then I still learnt a lot about
various relevance ranking algorithms while writing it :)

But that's the way that anything happens in SRW/CQL.  If enough people use
it then it might go into a future version of the standard without needing
an extraData space. But TBH, I don't think that will happen as there's a
Lot of more generally useful fields that would be included before this.

Compared to the Z39.50 model of 'If you show up to a ZIG, we'll put it
into the spec, kk thx bye' (appropriate apologies), this ensures that only
implemented and worthwhile extensions make it into the base spec, leaving
a very functional but still sleek standard.

Rob

--
      ,'/:.          Dr Robert Sanderson ([log in to unmask])
    ,'-/::::.        http://www.o-r-g.org/~azaroth/
  ,'--/::(@)::.      Special Collections and Archives, extension 3142
,'---/::::::::::.    University of Liverpool
____/:::::::::::::.
I L L U M I N A T I  L5R Shop: http://www.cardsnotwords.com/