Re: Rework the way multixact truncations work

On Tue, Dec 8, 2015 at 6:43 AM, Andres Freund <andres(at)anarazel(dot)de> wrote:
> On 2015-12-04 21:55:29 -0500, Noah Misch wrote:
>> On Thu, Dec 03, 2015 at 07:03:21PM +0100, Andres Freund wrote:
>> > Sorry, but I really just want to see these changes as iterative patches
>> > ontop of 9.5/HEAD instead of this process. I won't revert the reversion
>> > if you push it anyway, but I think it's a rather bad idea.
>>
>> I hear you.
>
> Not just me.
>
>> I evaluated your request and judged that the benefits you cited
>> did not make up for the losses I cited. Should you wish to change my mind,
>> your best bet is to find defects in the commits I proposed. If I introduced
>> juicy defects, that discovery would lend much weight to your conjectures.
>
> I've absolutely no interest in "proving you wrong". And my desire to
> review patches that are in a, in my opinion, barely reviewable format is
> pretty low as well.

I agree. Noah, it seems to me that you are offering a novel theory of
how patches should be submitted, reviewed, and committed, but you've
got three people, two of them committers, telling you that we don't
like that approach. I seriously doubt you're going to find anyone who
does. When stuff gets committed to the tree, people want to to be
able to answer the question "what has just now changed?" and it is
indisputable that what you want to do here will make that harder.
That's not a one-time problem for Andres during the course of review;
that is a problem for every single person who looks at the commit
history from now until the end of time. I don't think you have the
right to force your proposed approach through in the face of concerted
opposition.

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