Re: beta to release

On Sat, May 8, 2010 at 6:34 AM, Simon Riggs <simon@2ndquadrant.com> wrote:
> I've often faced the issue you describe. I think its difficult for
> everybody to help at this stage. In many ways it is a serialization and
> it's good that Tom holds the gate tighter than normal at this point.
>
> The main thing I've tried to do was map out plans for my time so you
> know which features will be worked on, in which order. Most of that
> planning needs to happen quietly because its not really possible to say
> "I intend to work on X" without it becoming a discussion about what X
> should look like, which is distracting to the release process.
>
> Having said that, I've often found that discussing things off-list with
> other hackers leads to later grief. I think Bruce wrote a good piece on
> that once. What 2 or 3 people think is a good way forwards is seldom
> shared by others, and many ideas fall foul of complete blockers, or
> simply easier or better ideas.

Yeah, I think we need to start having those discussions on-list.
Trying to develop things quietly so it doesn't become a distraction
can result in "wasting a lot of time going down a dead end".  Tom and
Alvaro saved me a ton of time on the
temporary-relations-get-different-filenames patch I submitted earlier
this week, and I really appreciate that.  My guess is that it didn't
take them much time to respond to my emails, either.  Even if neither
of them actually read for several months, I have a lot more confidence
that the basic approach is sound than I would otherwise, and I was
able to develop it much more quickly than would have been possible in
a complete vacuum.  Of course, as you say, we don't want to go nuts
and get into long arguments about things, but I think that high-level
design discussions should be viewed as in order.  This is important
(1) to avoid duplication, (2) to enable people to get done the work
their employers are willing to fund at the time their employers are
willing to fund it, and (3) to create a reasonable hope of some of the
big patches landing earlier in the cycle.

> Last year we both worked on the same issues. I'd ask that if you intend
> to work on anything you know myself or other hackers are working on,
> please ask so we can avoid duplicating any efforts.

I will try to do that.  Currently, the things that I know I will be
spending some effort on for 9.1 are (1) global temporary and/or
unlogged tables, (2) inner join removal, and (3) mentoring the GSoC
projects on matviews, json, and merge.  Everything else is pretty
amorphous at this point, but I typically send out design emails fairly
early on in the process.  Please also share your plans for 9.1.

Come to think of it, I think we should start a page on the wiki for
developers to list major projects they plan to work on for 9.1, maybe
also with a space to indicate whether the project is possible or
definite.

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