Emmanuel Lecharny wrote:
On 9/1/11 9:25 PM, Michael Ströder wrote:
> Emmanuel Lecharny wrote:
>> What kind of deficiencies have you found that make the original PSearch spec
>> unusable ?
> psearch is completely state-less. If a connector using it does not work it
> will miss deletes. Also restarting the connector will return *all* entries as
> results in a first phase. Your connector will not know when this first phase
> ends until it reaches the first result with ECN control attached.
Thanks. Yeah, but it's much more a lack of features than deficiencies.
For instance, a client requesting a PSearch may expect to receive
again, in order to load a local storage.
But yes, PSearch could be greatly improved. And the syncrepl mechanism could
be a good starting point for a better PSearch...
It does not make sense to improve draft-ietf-ldapext-psearch since the server
vendors who already implemented the current (expired) version of the I-D won't
modify their implementations. And any client/connector implementor who wants
to access OpenLDAP should implement syncrepl.