On Thu, Dec 08, 2011 at 09:29:28AM -0600, Merlin Moncure wrote:
you have a couple of very good options to achieve the same in postgres
-- pgbouncer, pgpool.
One of the central issues that Postgres has in the "enterprise" land
is exactly this sort of answer: "you have good options" but they're
"not part of the core release".
Now, we all know that this is a stupid and wrong way of thinking about
it. But one cannot complain about being held to those sorts of
enterprisey standards when one is having an enterprisey discussion.
The original analysis (on the blog) seemed to be primarily aimed at
exactly that sort of discussion, and I suspect that this is the kind
of thing that was meant by the "community leadership" not having
covered itself in stewardship glory. One of the "stewardship" tests,
from a business-analysis point of view, is whether you're going to be
able to find a reliable supply of experienced admins at all levels for
Having a bunch of different, indifferently-documented projects that
are all doing similar but slightly different things is, to someone
looking from that point of view, a liability and not a strength. I
happen to disagree, but it always seemed to me that something the
Postgres community did poorly (and I count myself in that number,
though less now than in the past) was understanding the hardships of
the integrator and coming up with reasonably simple answers for those
kinds of questions. It is not unreasonable to say that there are no
simple answers here; but as unhappy as it makes me, those reasonably
simple answers are necessary for some classes of users.
And let's face it: companies like Oracle (and products like MySQL) are
in a position to treat those sorts of answers as part of the cost of
doing business, because they have revenue associated with their
licenses so they can pay for coming up with those answers that way.
In Postgres-land, everyone needs to charge money for those answers
(i.e. be consultants), because that's the only real place to make a
living. Alternatively, you can put together those answers as part of
your own package; but in that case, it's not "the core PostgreSQL
product", but something else.
In this respect, the decision of the core team a number of years ago
to say, "We're going to have 'integrated' replication that does x, y,
and z," was the right one, despite the fact that it undermined the
momentum of other interesting projects (and ones better suited to some
environments). Sometimes, it's better to cut off options.