12 Nov
2009
12 Nov
'09
12:11 p.m.
On Thu, Nov 12, 2009 at 12:29, Martin Aspeli <optilude+lists@gmail.com> wrote:
Leonardo Rochael Almeida wrote:
Wouldn't it be better to just move IPubFailure before the abort? [...] I can see the usefulness of the transaction being already doom()ed before this event, but not aborted.
There is an IPubSuccess which runs after the transaction is closed. Doing post-processing after closure may be useful to keep transactions shorter and thus reduce the risk of conflict errors. I don't think an extra event will cost us much, so being consistent and granular is probably best.
Fair enough. I still suggest doom()ing the transaction before this event.