1 |
On 1 July 2013 19:16, Chí-Thanh Christopher Nguyễn <chithanh@g.o> wrote: |
2 |
> Rich Freeman schrieb: |
3 |
>>> I did not say that maintainers can ignore policy. The removal of |
4 |
>>> ebuilds must follow certain rules which are set in policy. |
5 |
>> IMO this really doesn't say anything at all. Of course maintainers |
6 |
>> have to follow policy. The question is what should the policy be? |
7 |
> |
8 |
> I think I am beginning to understand what you want to say. You want the |
9 |
> council (or some other body) to act as enforcer to bring maintainers in |
10 |
> line with prevailing opinion. Sorry, I am not supporting that. |
11 |
> |
12 |
>> The original question was whether we should be forking or splitting |
13 |
>> packages over adding single files when the maintainer doesn't want |
14 |
>> them in the original package. Right now we have no explicit policy |
15 |
>> governing this. |
16 |
> |
17 |
> Yes, and due to the absence of policy it is the maintainer who has the |
18 |
> final say. That is how it is currently, and how I think it should be in |
19 |
> the future too. GLEP 39 specifically says that you cannot monopolize the |
20 |
> package, and anyone can fork the ebuild. That is sufficient remedy |
21 |
> against uncooperative maintainers in my eyes. |
22 |
> |
23 |
> I wrote in a previous message that I would reconsider my position if the |
24 |
> amount of forked packages grew to unmanageable proportions. But until |
25 |
> then, if the kids cannot get along they shall play on separate playgrounds. |
26 |
> |
27 |
>>> With x32 specifically, a number of people including some upstreams think |
28 |
>>> that the whole concept is a bad idea. A case could be made for patches |
29 |
>>> that #ifdef x32 and which compile to a no-op on other arches, but even |
30 |
>>> those must be maintained. What if the patch no longer applies after a |
31 |
>>> version bump? |
32 |
>> Well, since I'm only talking about WELL-SUPPORTED project-related |
33 |
>> work, just ask the project team to fix the patch. If they don't in a |
34 |
>> reasonable timeframe, then it isn't well-supported and the maintainer |
35 |
>> can do whatever they want with it. Project teams should only take on |
36 |
>> patches if they think they can sustain them. Most likely for |
37 |
>> something like x32 they'd only do it for strategic packages anyway |
38 |
>> (toolchain, etc). |
39 |
> |
40 |
> But in this hypothetical scenario you have unloaded additional work on |
41 |
> the maintainer. He just wants to bring the latest and greatest version |
42 |
> to his users, and the failing patch prevents him from doing that. He has |
43 |
> to request and then wait for the a new patch, and if he bumps the |
44 |
> package anyway after timeout risks breaking x32 users' systems. |
45 |
> If he does this voluntarily, fine. If the patch was forced on him, |
46 |
> putting him in this dilemma is not fair. |
47 |
> |
48 |
> |
49 |
> Best regards, |
50 |
> Chí-Thanh Christopher Nguyễn |
51 |
> |
52 |
> |
53 |
|
54 |
Thank you for this voice of reason. |
55 |
|
56 |
-- |
57 |
Cheers, |
58 |
|
59 |
Ben | yngwin |
60 |
Gentoo developer |