> Using pg_control to get checkpoint position speed up things but
> to handle possible pg_control corruption we obviously should
> implement reading existent log segments (from the last one -
> newest - to oldest) to get last checkpoint.
I think this would be *very* important :-)
Andreas
