>>> [log in to unmask] 20-4-04 19:46:25 >>>
>to produce consistent, useful results across a range of servers. It
>allow us to produce applications that *looked* useful on the surface,
>whenever you'd try to do somethind serious with them, you'd start to
>realize that the results were completely unpredictable and rarely bore
>relationship to what you would get from searching the native
>the same resources.
In that case it wasn't done right then. What I propose should allow
responses that bare very much relationship with the native interface and
be very predictable but then according to a very well defined scheme.
The only difference is that when a server does not support this, it just
returns the conventional response. I do not consider different the fact
that different servers have capabilities as unpredictable.
>SRW potentially makes the job of building and operating metasearch
>applications a LOT simpler (and less expensive!) by reducing the need
>'screen-scraping' or proprietary APIs.. however, it doesn't, and
>hide the fact that different resources have different contents. I
>want to use a service like the one you describe in a metasearch
I agree, except for the fact that the service that I describe does
exactly what you want: it does not hide the differences. The fact were
we possibly might disagree is that I want to see the differences in the
response and rather than having to deal with them in the request.
Currently the user is very much dependend of what a metasearch portals
or gateways offer and will in many cases not see the differences
because of the common denominator approach: only send requests that can
be sent to all.
Clients have to be intelligent to deal with differences in requests
than to accept extra XML blocks that are just shoved into an xslt
processor and are just neglected when not supported.
I think that one very important assumption in my proposal is the use of
SRU in a user or institution controlled environment with personalised
stylesheets to process the responses.