[Bizgres-general] SetBufferCommitInfoNeedsSave

[Jim C. Nasby]

I seem to recall some discussions in the past about vacuuming pages as
they were written to disk. It seems like also setting the info bit would
be a win (though of course this wouldn't do any good for really large
data loads).

BTW, this isn't mentioned at all in
http://www.postgresql.org/docs/8.0/interactive/storage-page-layout.html.

On another note, if setting of this bit was WAL logged, would it then be
possible to remove t_xmin and t_cmin from tuples that have the bit set?
ISTM that saving 8 bytes per tuple would be well worth it, since we have
to re-write the page anyway. I guess this would depend on there being no
older transactions running when the bit was set.

I'll add this to my mental list of possible ways to reduce tuple header
size.
-- 
Jim C. Nasby, Database Consultant               decibel at decibel.org 
Give your computer some brain candy! www.distributed.net Team #1828

Windows: "Where do you want to go today?"
Linux: "Where do you want to go tomorrow?"
FreeBSD: "Are you guys coming, or what?"