drizzle-discuss team mailing list archive
Mailing list archive
Re: tables must have a primary key
On Wed, 7 Apr 2010 14:41:26 +0200, Arjen Lentz <arjen@xxxxxxxxxxxxx> wrote:
> From observation, the appear to be other consequences to not
> specifying a primary key for InnoDB, in terms of locking behaviour and
> So the implicit rowid/pk thing is, apparently, not quite equivalent.
my auto-increment implementation is quite different than the innobase
Since we don't at all care about or support statement based replication,
we can be a lot more sane in locking (or lack thereof).
Although I'm not really caring about performance yet... in theory the
embedded_innodb auto increment implementation should be much better.