On 2018-04-07 13:33:53 -0400, Bruce Momjian wrote:
> To summarize how I see this patch, we have this workflow at the top of
> the TODO list (which I think Simon helped with or suggested):
>       Desirability -> Design -> Implement -> Test -> Review -> Commit
> I think the MERGE patch spent a long time getting through the first and
> second stages

The current implementation effort publicly started 2017-10-27. For a
major feature that's really not that long ago. There also were a few
gaps in which no public development/discussion happened.

>, and now the objections appear to be related to implementation.  I
>think the implementation issues only appeared during the final
>commitfest, which made it feel like a new patch.  Yes, it had been
>through the first two stages before the final commitfest.

I'm not sure there was agreement on the design even. A lot of that has
been discussed until very recently.

> I think one of the missing rules we have is that when we say no new
> large patches in the final commitfest, do we mean that all _three_
> stages should be solidified before the final commitfest?  I have never
> been clear on that point.

I think the implementation at the least should be roughly ready, and
implement a roughly agreed upon design. It's fine to change things
around, but major re-engineering surely is an alarm sign.


Andres Freund

Reply via email to