[Date Prev][Date Next]
RE: status of multi-master replication?
> -----Original Message-----
> From: owner-openldap-software@OpenLDAP.org
> [mailto:owner-openldap-software@OpenLDAP.org]On Behalf Of John Madden
> Thanks - that provides HA without load balancing. That may be sufficient
> for our needs, but with multimaster, we'd be able to load balance
> everything as well. However, actually *compiling* openldap with
> multimaster is another issue entirely. (I'd like to be able to
> ./configure --enable-multimaster; make; make install, but no, I've got to
> modify configure.in and run autoconf, which fails miserably... quite a
> Upon further archive searching, I found mentions of multimaster as far
> back as 1999. Why is it that a change which only effects 6 or 7 source
> files with minimal changes takes roughly 3 years to become "not
> experimental?" The concept itself even seems fairly simple, so why is
> this so difficult?
OpenLDAP is volunteer-driven; if something doesn't get completed that means
nobody stepped in to complete it. This concept seems fairly simple too, so
why do people continue to complain about missing features?
If you believe it's so simple, and find it such a nuisance that it's not
done, why not do something productive about it?
I can't speak for anyone else, but the feature has never even remotely been
a priority for me personally. I believe, if you actually spend some "quality
time" thinking about it, you'll encounter an insurmountable problem -
resolution: If two (or more) changes are made to the same object on different
masters, there are several conflict scenarios that cannot be automatically
resolved, especially if the object contains attributes that don't specify an
Equality matching rule. These conflicts are impossible to resolve, some are
even impossible to detect, giving you a database of unknown integrity, which
to me is completely useless.
-- Howard Chu
Chief Architect, Symas Corp. Director, Highland Sun
Symas: Premier OpenSource Development and Support