In case of SRU via a webserver the request wil not even reach the
SRU-service: the webserver will give the error message.

Theo

>>> [log in to unmask] 12/16 5:59  >>>
> Date: Tue, 16 Dec 2003 11:30:41 -0500
> From: "LeVan,Ralph" <[log in to unmask]>
>
> The problem with returning the 404 is that there is no option for
> telling them what went wrong.  In my case, I know that the database
> name part of the URL didn't match anything I had and I'd like to
> tell them that.

In other words, what you need is a Grown Up Information Protocol that
has notions like "database".  But that's precisely the kind of
"unnecessary complexity" we traded away in the move from Z39.50 to
SRW/U.  In the SRU/W world, there really is no such thing as a
database: just servers.  Some URLs have SRW/U servers on the end and
some don't.

I think the best we can do is pick one of the less frequently used of
the more-or-less appropriate HTTP error codes and use that.  Which may
be pragmatically useful.  But conceptually, 404 is right.

 _/|_
_______________________________________________________________
/o ) \/  Mike Taylor  <[log in to unmask]>
http://www.miketaylor.org.uk
)_v__/\  "What would the Coliseum have looked like if Liverpool city
         council been in charge?" -- Paul Clough.

--
Listen to my wife's new CD of kids' music, _Child's Play_, at
        http://www.pipedreaming.org.uk/childsplay/