[Nix-dev] Keeping nixpkgs up to date

[Michael Raskin]

>On Fri, Aug 29, 2014 at 3:59 PM, Michael Raskin <7c6f434c at mail.ru> wrote:
>>
>> I would say that ZHF has some progress specifically because it is about
>> reasonable efforts to improve signal-to-noise.
>>
>> If we want packages to get updated, we need to merge pull requests
>> faster and not let pull requests linger in limbo. That would make
>> sending PRs for updates look like a good idea (so less burden on
>> maintainers to monitor updates). For complex questions, we should Cc:
>> the maintainers of the packages in question (or of the most similar
>> packages with maintainers) — but the idea is that more people should
>> read pull requests. Once in a couple of weeks I find time to reduce the
>> amount of pending PRs by a page. Five more people among committers to do
>> the same and we'll have no pending PRs without good reason for
>> a detailed discussion, I hope…
>>
>
>Well, I'm not the committer. But I think the problem with looking to
>PR, is that there are 110 unsorted PRs. It's hard to look them all in
>one go. And it's useless that all committers look at them all. If
>there would be 11 groups of 10 PRs, it's easier to look at 10 PRs
>several times a week.

This is not how you solve the problem! You look at 20 random PRs, merge
or ask someone to merge 10 of them, propose rejection of a couple, and 
CC a couple people you think could know something for eah of the rest…

After fifteen passes the problem gets reduced to a manageable size.

Look at ZHF: there was nothing bout division, ther was just a plea for
attention to a problem. When the problem got reduced, some natural 
groups of broken packages emerged that were dispatched by those who 
understood them.

Most of pull requests are simple to review as net-positive…