On Sun, Sep 16, 2007 at 08:04:06PM -0400, Jan Wieck wrote:

> achievable as cheap as you think. One would first have to agree that all 
> nodes in the whole cluster have to have tables containing the same 
> columns. Then, the column names can be mapped to short integers in a new 
> slony configuration table. 

So, let's stop there, please, because this seems to me like a
regression.  The current arrangement may allow us flexibility that we
would lose under such a proposal.

The goal of the OP seems to be "optimization" in exactly the way that
MySQL construes it: for some subset of database-like operations, it
is fast.  But we are working on a general purpose tool.  You have to
weigh the costs of losing generality, too.  If you want a bespoke
tool that replicates perfectly in a way that is ideal for your
particular use pattern of the database, then you'd best hire some
developers for your own development, rather than ask the wider world
to optimise for your own application.

A

-- 
Andrew Sullivan  | ajs at crankycanuck.ca
The plural of anecdote is not data.
		--Roger Brinner