RE: Multimaster further work

[Derek Simkowiak <dereks@itsite.com>]

> such, multi-master replication offers no benefit, while introducing a slew of
> potential problems.

	On this point, I disagree.  The goal of loadbalancing is not just
to distribute load; it's also to provide high availability and ease of
management and maintenance.

	It's far easier to manage a set of homogenous master nodes in a
cluster than it is three separate configurations: master-primary,
master-failover, and slave.  It's that versus just "multimaster".

	Furthermore, in the load-balancer, there are more moving parts
with separate configs.  You either need a unique I.P. Address or port for
the master, if you have a write-LDAP seperately from readonly-LDAP slaves.
And then you need to configure the loadbalancer for failover on LDAP slave
(with no "failover" at all required for multimaster).  Don't forget to
configure your LDAP client apps differently, depending on if they are
allowed to write or not.

	Compare that to just simple loadbalancing on a single ip/port pair
-- with the added benefit of high availability (which a single master does
not offer).

	And, I can gracefully fail out my homogenous nodes for upgrades,
one by one, without any interruption of service to the customer.  A
master-failover config require stopping the service -- even if only for a
second -- and that is something that always Looks Bad, no matter how
incidental.

	Also, if the master goes down at 3am, you now have a single point
of failure.  With 6 masters (i.e., each node is a master), you still have
5 nodes that will service customers.

	Not to mention the locality advantage (if you have disparate
offices spread across the globe).  The ping to the (single) master may
great for New York, but it will sure suck trying to use that (single)
master from Tokyo.  It's all in the background if you also have a master
in Tokyo.

	Even if these benefits are relatively arbitrary, they give the
customer ease of mind.  Given the choice, I'd want my infrastructure built
around multiple masters instead of a single failover config.

> In any profession, the percentage of people proficient at thinking
> critically in that field is a tiny amount compared to the field as a

	I disagree and find this to be a rather elitist viewpoint.  I
understand that most people act like hairless monkeys but I still find
great value in their experience and viewpoints.  I'd love to go into to
"why" but it's offtopic and rather philosophical.  Email me privately if
you feel the need.

> I also think it's naive to believe that developers are not also
> endusers themselves, and fully cognizant of the features needed to
> apply an arbitrary technology to a real world problem.

	I do not "naively" doubt the developer's cognition.  I just think
10 developers will not be as creative (or critical) as 100 endusers, some
of whom have developer experience.  Also, the way they model the "real
world problem" may be completely different than the way someone else
models it.  (This closely relates to my statement above.  I'd like to get
off this topic and get back to the issue of multimaster conflict
resolution.)


--Derek