"Tom Lane" <tgl@sss.pgh.pa.us> wrote
>
> Hm, I already talked about that once:
> http://www.postgresql.org/files/developer/transactions.pdf
> but perhaps that's not the level of detail you are after?
>
Yeah, I've read the presentation -- and yes, that's not the level I am
after. Actually, I guess the completeness problem can be clarified with a
table:
a_complete_and_clear_division_of_transaction_time
SatisfySnapShot Y Y N N ...
SatisfyVacuum 1 2 2 3 ...
Not sure how to give the corrctness proof -- maybe if the
"a_complete_and_clear_division_of_transaction_time" is designed well enough,
we can find some consistency?
And as Alvaro suggested, maybe that's too narrow topic for a presentation --
so if the above idea is in the right track, I'd like to write an initial
document (so you guys can modify it) if nobody is interested in doing that.
Regards,
Qingqing