Simon Riggs <simon(at)2ndQuadrant(dot)com> writes:
> On Mon, 2009-12-07 at 19:07 +0200, Heikki Linnakangas wrote:
>> Why not just follow the example of postresql.conf?
> Much better idea.
Rather than reinventing all the infrastructure associated with GUCs,
maybe we should just make the recovery parameters *be* GUCs. At least
for all the ones that could be of interest outside the recovery
subprocess itself.
As an example of the kind of thing you'll find yourself coding if you
make an independent facility: how will people find out the active
values?
regards, tom lane