It is absurd to exclude multi-master
replication from normal LDAP configurations. Every large deployment
uses it, because they can't tolerate a single point of failure. That being
the case, there is no consensus not to use it. Just the opposite:
There is a consensus that it must be used.
One can support multi-master replication
and the increment operation. However the increment operation, and
all operations on that attribute value, must occur on a particular master,
and must be deferred if that master is not available.
John McGarvey
IBM directory architect
919-877-4892 or t/l 254-4892
"David Boreham"
<david_list@boreham.org> Sent by: ldapext-bounces@ietf.org
> I hope we can avoid rehashing this. I think
consensus is
> pretty clear on the matter. X.500 specifications and
> implementations presume that one and only one DSA is
> authoritative for an entry held in the Directory.
I see, you're really arguing from the perspective
of how the service is implemented rather than any
specific service semantics as seen by clients.
So it's simply enough to say that MMR isn't allowed,
and not important to say _why_.
I get your thinking now, thanks.
_______________________________________________
Ldapext mailing list
Ldapext@ietf.org https://www1.ietf.org/mailman/listinfo/ldapext