I agree completely. I was not referring to a particular case, but there
are some cases where I would have liked a lower implementation barrier
like for example distributing a single user query to heterogenious
services and letting the services respond with the best they can offer.


Theo


>>> [log in to unmask] 28-6-04 12:01:06 >>>
> I am in favour of lowering the implementation barrier for
> clients  but I don't think that is the same as encouraging
> people  to write bad clients. Maybe it helps bad client
> writers to write better clients -:) Increasing the
> implementation barrier makes it more difficult to write good
clients.

We agree with that - however there are certain things which whilst
lowering the barrier are also "bad"

For instance:

i) using XML rather than BER is a good way of lowering the
implementation barrier

ii) insisting (or implying) that the namespace prefix for SRW is
always
srw so that a client can do regular expression matching on
srw:searchRetreiveResponse rather than looking for and using the
prefix
defined in an xmlns: also lowers the implementation barrier but is
"bad"

Matthew