Re: vacuum locking

Am Donnerstag, 23. Oktober 2003 15:26 schrieb Tom Lane:
> ... if all tuples are the same size, and if you never have any
> transactions that touch enough tuples to overflow your undo segment
> (or even just sit there for a long time, preventing you from recycling
> undo-log space; this is the dual of the VACUUM-can't-reclaim-dead-tuple
> problem).  And a few other problems that any Oracle DBA can tell you about.
> I prefer our system.

of course both approaches have advantages, it simply depends on the usage
pattern. A case where oracle really rules over postgresql are m<-->n
connection tables where each record consist of two foreign keys, the
overwrite approach is a big win here.