"Scott Marlowe" <smarlowe@qwest.net> writes:
> On Sun, 2004-07-25 at 22:23, Tom Lane wrote:
>> I don't think
>> it's either practical or interesting to try to introduce an equivalent
>> layering into Postgres.
> I can possibly see a use for a row locking storage system, i.e. non MVCC
> for some applications. But I can't see it being worth the amount of
> work it would require.
I can't either. The implications of such a thing really are so
far-reaching that it could not be isolated in a nice little layered API.
To take one example, we'd have to reexamine the locking and crash-safety
behavior for every single command that updates the system catalogs.
For better or for worse, I think we're married to MVCC.
regards, tom lane