>
> * Consider transaction start/end performance improvements
>
>   http://archives.postgresql.org/pgsql-hackers/2007-07/msg00948.php
>   http://archives.postgresql.org/pgsql-hackers/2008-03/msg00361.php