Matthew J. Dovey wrote:
> I'm a little worried about how we are overloading this single endpoint
> (especially in the case of Yaz doing inspection to work out whether
> protocol is http or z39.50, which admittedly is "clever").

Well, yeah -- we don't _force_ anyone to use this functionality!  :-)

> So if adopted in SRU/W, we would publish the Explain endpoint, and that
> would give us whether we supported SRU (and its endpoint), SRW (and its
> endpoint), Update (and its endpoint), etc...

Yep, that's all sound.

> If we are going to introduce SRU POST, I think I would suggest we do
> this (although I would anticipate SRU GET and POST being always on the
> same endpoint). It is backwards compatible with the current situation in
> that all of these endpoints could be the same (if you want to do clever
> inspection of the incoming data to workout how to handle it), but would
> allow multiple endpoints for those who might regard having to do this
> inspection as an unexceptable overhead.


Agree 100%.  My reference to what our own implementation was meant only
to be a proof-by-existence of the feasibility of that approach -- not a
demand that everyone else implement the same approach.