[ZODB-Dev] [Ann] "ReadCommitted" transaction isolation level

[Dieter Maurer]

Hi Shane,

Shane Hathaway wrote at 2003-3-22 02:45 -0500:
 > ...
 > Enough rambling. ;-)  The paper is here:
 > 
 > http://citeseer.nj.nec.com/berenson95critique.html

Thank you for the reference.

Reading the article, I recognize that the patch
implements "Snapshot isolation" (modulo some ZEO race condition
where invalidation messages arrive too late, this can be fixed
by syncing at transaction start).

Snapshot isolation is stronger than "Read committed" but
weaker than "Repeatable read".


Dieter