Re: ALTER TABLE lock strength reduction patch is unsafe

On Fri, Jun 24, 2011 at 5:28 PM, Simon Riggs <simon(at)2ndquadrant(dot)com> wrote:
> I've explained all of the above points to you already and you're wrong.

We're going to have to agree to disagree on that point.

> Do you advocate that all ALTER TABLE operations use
> AccessExclusiveLock, or just the operations I added? If you see
> problems here, then you must also be willing to increase the lock
> strength of things such as INHERITS, and back patch them to where the
> same problems exist. You'll wriggle out of that, I'm sure. There are
> regrettably, many bugs here and they can't be fixed in the simple
> manner you propose.

I think there is quite a lot of difference between realizing that we
can't fix every problem, and deciding to put out a release that adds a
whole lot more of them that we have no plans to fix.

> It's not me you block Robert, I'm not actually a user and I will sleep
> well whatever happens, happy that I tried to resolve this. Users watch
> and remember.

If you are proposing that I should worry about a posse of angry
PostgreSQL users hunting me down (or abandoning the product) because I
agreed with Tom Lane on the necessity of reverting one of your
patches, then I'm willing to take that chance. For one thing, there's
a pretty good chance they'll go after Tom first. For two things,
there's at least an outside chance I might be rescued by an
alternative posse who supports our tradition of putting out high
quality releases.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company