1 |
Andreas K. Huettel: |
2 |
> |
3 |
> Doing a thorough review of submitted code can be harder than writing it |
4 |
> yourself. |
5 |
|
6 |
Only if the PR is low quality and I don't know the contributor yet. It |
7 |
could be simplified with automatic travis build logs etc. |
8 |
That has already been done in other distros and is nothing spectacular. |
9 |
|
10 |
> If I merge, I'm responsible for it. Which is why I am not entirely |
11 |
> happy with a pull request based workflow... To put it blunt and exaggerate, |
12 |
> why should I put in a lot of work and take over responsibility for a bunch of |
13 |
> lazies who want to contribute somehow but feel disinclined to do it properly? |
14 |
> |
15 |
|
16 |
Because you want to teach them? |
17 |
|
18 |
> The recruiting is the process where people prove they can themselves take |
19 |
> responsibility for our users and are willing to do it. |
20 |
> |
21 |
|
22 |
You prove once and then are free to mess up regularly? Hmm. |
23 |
|
24 |
>> I've been asked recently by a community member if he should bother to |
25 |
>> try to become a gentoo dev, because he heard it's a pain. |
26 |
>> Then again... regularly contributing to gentoo via bug reports is even |
27 |
>> more a pain... so he's just running yet another personal overlay. |
28 |
> |
29 |
> Yes. Exactly. 100% true. We all know this situation. |
30 |
> |
31 |
> So, as you point out, a decentral operation is one solution (reduced core |
32 |
> repository, more overlays). What it means though is that all quality control |
33 |
> measures in place (security team, arch team, qa team, rules what needs |
34 |
> discussion on the lists, ...) only apply to the main tree, and the overlays |
35 |
> slip out of it. Which is something that I plainly do not want. Even if things |
36 |
> do not work perfectly now, we'd still lose far too much. |
37 |
> |
38 |
|
39 |
Please don't mix my idea of decentralization with an enforced review |
40 |
workflow. Both concepts are totally orthogonal. You can have a |
41 |
centralized concept and very good review and a decentralized concept |
42 |
with very poor review and any other combination. |
43 |
|
44 |
I'd still vote for combining this with a decentralized concept, because |
45 |
it may solve even more problems, but that point is irrelevant for this |
46 |
topic. |
47 |
|
48 |
> Here's my answer to your contributor: "It's great that you're considering it, |
49 |
> and we're happy to help you. It's worth becoming a full dev, since you have |
50 |
> far more possibilities to directly contribute and also influence where Gentoo |
51 |
> is going. However, it's going to be some work and it will need some |
52 |
> commitment." |
53 |
> |
54 |
|
55 |
I already told him to focus on contributing to project overlays if he |
56 |
cares about ebuild writing only... so that he gets reviews and doesn't |
57 |
just stick to his own overlay. |
58 |
|
59 |
>> The fact that projects like proxy-maintainers and sunrise have to exist |
60 |
>> are proof that we are doing something wrong! I'm not discrediting any of |
61 |
>> those. They fill a gap, but the question is... why do we have that gap |
62 |
>> in the first place? |
63 |
> |
64 |
> Because you can't answer the quizzes in 166 characters and submit them via |
65 |
> Instagram? |
66 |
> |
67 |
|
68 |
I was referring to missing tools, workflow and review culture. Git might |
69 |
improve it... but it's not enough. See sunrise... the migration to git |
70 |
didn't add anything of value. |
71 |
|
72 |
|
73 |
Because contributing is cumbersome we need projects like |
74 |
proxy-maintainers. But from my experience that doesn't work very well. |
75 |
Project overlays on github currently work better. Less indirection. |
76 |
|
77 |
And I've sometimes seen people in the git log who have been contributing |
78 |
for ~3+ years with high-quality ebuilds without ever being asked to join |
79 |
the team. |
80 |
Then I asked them via mail and now guess the answer... (and... is it |
81 |
even bad?) |