On Sat, Oct 15, 2005 at 05:58:20PM -0700, Chris Travers wrote:
> Repeat after me... Data integrity, like security, is a process, not a
> product.
I understand that. The problem is not that _I_ don't understand it,
but that the market for real, industrial multi-master replication is
so far relatively small, and it looks like the market for RAC. Which
means that one has to present the complete package to people who
don't know what a Postgres is, and don't want to.
It is _of course_ not impossible to provide for this. But one needs
canned, "this is standard" ways to do these things, or else the
people who are in a position to authorise Postgres use instead of
something else are not going to do it. They don't want to be guinea
pigs, and they can afford not to be. (Note that I'm not suggesting
this is really a technical problem. It's just a social problem to
which we have to present technology.)
The sad truth of the matter is that if you want to alter the social
circumstances in favour of some new, unconventional approach, your
unconventional approach has to be _better than_ the existing
convention, not merely as good as. And in the market we're talking
about "cheaper" is not the main consideration for "better than". I
think other arguments are useful -- access to source (and therefore
auditability) is an obvious one -- but one needs to establish a
well-known set of practices around these things if one wishes to be
taken seriously for this kind of application.
A
--
Andrew Sullivan | ajs@xxxxxxxxxxxxxxx
In the future this spectacle of the middle classes shocking the avant-
garde will probably become the textbook definition of Postmodernism.
--Brad Holland
---------------------------(end of broadcast)---------------------------
TIP 9: In versions below 8.0, the planner will ignore your desire to
choose an index scan if your joining column's datatypes do not
match
