On 11/06/2011 06:00 PM, Tom Lane wrote:
> Peter Geoghegan<peter@2ndquadrant.com> writes:
>
> > A major consideration was backwards compatibility;
> This is not a consideration that the community is likely to weigh
> heavily, or indeed at all. We aren't going to back-port this feature
> into prior release branches, and we aren't going to want to contort its
> definition to make that easier.
Being able to ship a better pg_stat_statements that can run against
earlier versions as an extension has significant value to the
community. Needing to parse log files to do statement profiling is a
major advocacy issue for PostgreSQL. If we can get something useful
that's possible to test as an extension earlier than the 9.2
release--and make it available to more people running older versions
too--that has some real value to users, and for early production testing
of what will go into 9.2.
The point where this crosses over to requiring server-side code to
operate at all is obviously a deal breaker on that idea. So far that
line hasn't been crossed, and we're trying to stage testing against
older versions on real-world queries. As long as it's possible to keep
that goal without making the code more difficult to deal with, I
wouldn't dismiss that as a useless distinction.