1 |
On Sat, Mar 29, 2014 at 10:36 AM, William Hubbs <williamh@g.o> wrote: |
2 |
> |
3 |
> We are supposed to be a team, and if a few people aren't team players, |
4 |
> I would rather see energy spent on fixing that issue instead of trying |
5 |
> to legislate every possible corner case. |
6 |
|
7 |
So, a few thoughts on the whole issue in general. |
8 |
|
9 |
1. When somebody wants to do something in a way completely different |
10 |
from how things have been done before or in a way that impacts how |
11 |
many packages will be maintained, it is best to discuss it on the |
12 |
lists first. That doesn't mean that you can't proceed without |
13 |
unanimous consent, but rather that even good ideas can become better |
14 |
ideas with more input. That said, when an issue comes up those who |
15 |
object need to at least speak up ahead of time, because after-the-fact |
16 |
isn't the time for I-told-you-so. That has come up a few times lately |
17 |
- you don't need to have the last reply on the thread to win the |
18 |
argument, but that doesn't mean that you shouldn't speak up at all. |
19 |
|
20 |
2. When there is a change in the tree that threatens to be disruptive |
21 |
to users/dev/etc QA has the right to step in. The necessity of doing |
22 |
this should really be considered first - if simply saying "hold on, |
23 |
don't do any more of this for a few days" is sufficient to get by, |
24 |
that should be preferred over masks/bans/etc. On the other hand, if |
25 |
users are being adversely affected by something and this will get |
26 |
worse with time, it is better to stop the bleeding. Think of it like |
27 |
asking a court for an emergency stay/injunction/etc - courts generally |
28 |
prefer to hear the whole case before taking action, but if there is |
29 |
some kind of irreversible harm at stake, they'll act first and sort it |
30 |
out later. Obviously QA should manage its own guidelines as to when |
31 |
individuals should act in the name of QA. |
32 |
|
33 |
3. In general the absence of an explicit policy should never be |
34 |
considered grounds for ignoring #1 or preventing #2. The real |
35 |
principle is cooperation and managing disruptive change. QA can step |
36 |
in even if you can't quote a specific rule that was broken (though |
37 |
they should be careful before doing so), and I really don't want to |
38 |
hear excuses like "well, there was no rule saying I had to play nicely |
39 |
with the other kids." We really don't want to have to maintain the |
40 |
Code of Gentoo Regulations. |
41 |
|
42 |
So, I'd certainly endorse that things like profile changes, eclass |
43 |
changes, and changes that impact many reverse-deps should be announced |
44 |
on the list and coordinated in general if they aren't routine. That |
45 |
said, I don't see the need to codify this in some kind of explicit |
46 |
policy - I'd argue that it already is policy. I don't want to get |
47 |
into back-and-forth on how to word a policy so that a dev masking |
48 |
their own package is fine but a dev removing bash from the system set |
49 |
isn't. If somebody can't tell the difference between these, they |
50 |
shouldn't have commit access. |
51 |
|
52 |
Rich |