Maybe I missed a part of this discussion, but I think we should try as
much as possible to apply a third criterium for accepting new features
in SRU/W and that is:
1) that when a client ask for this new feature a server that did not
implement this feature should be able to give a "reasonable" response -
not being an error message.
2) that when a server provides a new feature the client should be able
to neglect this when he is "not aware" of this technology.
In this way we improve backwards compatability and robustness.
I know that it will not always be possible to apply this criterium but
in this case it is.

Reasoning from a perspective of not having much control over the
clients or servers one is communicating with, we should try to avoid
implementation of new features that need negociations based on
diagnostics to let the client find out what the best question is he can
ask the server. The barrier for implementing this kind of functionality
is usually high. However - when new features are provided unsollicited
in a way they can be neglected - it will lower the barrier for adding
code to make use of such a new feature.

In the case of xPath there is an easy escape (not a diagnostic) for
servers that did not implement this by providing just ordinary captions
or complete records (whatever is the default) and returning the xPath
parameter as "unknownParameter".



>>> [log in to unmask] 8/5/03 7:24:11 nm >>>
> Here, we have a real implementor group, a real problem cited,
> and the person
> proposing the feature (Rob), I'm confident will implement it.
>  So I don't
> think the Z39.50 "hypothetical problem" argument applies.


I had no doubt that Rob would implement it - however, some of the
orphaned stuff in Z39.50 have been because only one person needed
implemented the feature and they subsequently moved on. So part of the
reason I questioned xPath was to bring people out of the woodwork to
ensure that we would have more that one person implementing it.

The other question I wanted to raise was whether we really needed to
SRW or whether there was a perfectly good way to do what Rob wanted in
another way with SRW as is.

I think these are two criteria we should judge new stuff in SRW by to
avoid orphaned technologies.

[And as said, I happy to accept from the discussion that Rob's Xpath
does pass both of these]