Re: pg_system_identifier()

> After someone in IRC asked if there was an equivalent to MySQL's
> server_id, it was noted that we do have a system identifier but it's not
> very accessible.
>
> The attached patch implements a pg_system_identifier() function that
> exposes it.

Would it make sense for such identifiers be standard UUID 
(http://en.wikipedia.org/wiki/UUID)?

Should there be a UUID per cluster? and/or per database, possibly deduce 
from the cluster one? Should it be configurable, say from 
"postgresql.conf"?
  get_pg_uuid()  get_pg_uuid('template0')

Note that there is a set of uuid functions provided as a module that may 
help.

-- 
Fabien.