On Wed, 2008-08-20 at 10:46 +0900, ITAGAKI Takahiro wrote:
> One thing to worry about is a confliction of RmgrId. We can check
> conflictions in redo because rmgrs are actually registered, but
> we might need to check conflictions even in a normal running.
> Extensions that write own XLog record can use arbitrary RmgrIds
> without restrictions.
That sounds quite hard. I'm putting in a check on the id itself, which
will be more than we had before...
-- Simon Riggs www.2ndQuadrant.comPostgreSQL Training, Services and Support