1 |
On 22.7.2022 22.54, Alec Warner wrote: |
2 |
> |
3 |
> I suspect this will end up being a process constraint, not a technical one. |
4 |
> |
5 |
> E.g. technically any dev can grant any contributor access to any |
6 |
> package (because managing thousands of fine-grained ACLs seems no |
7 |
> bueno to me.) |
8 |
> |
9 |
> This is fine (the same is true for anything else a developer does in ::gentoo.) |
10 |
|
11 |
So in theory yes. But you'd have to request the access from infra (and |
12 |
deal with gpg keys) which puts at least some sort of a barrier. |
13 |
|
14 |
In reality there's not too many of us who deal with user contributions, |
15 |
and I personally wouldn't start giving access to everyone who's ever |
16 |
contributed. Basically I was looking at these very simple packages that |
17 |
update sometimes multiple times a week, where the ebuild itself rarely |
18 |
changes. These are very "problematic" packages for us to handle because |
19 |
currently, and for a long time now, our "reaction" time is slower than |
20 |
upstream pushing updates. But I can see us extending this into trusted |
21 |
contributors being able to push and maintain new packages, which also |
22 |
have been a problematic issue for the proxy-maint project for a while now. |
23 |
|
24 |
|
25 |
> |
26 |
> So to clarify: |
27 |
> - We would like to make these people developers. |
28 |
> - Some don't want to be developers for various reasons (the quiz |
29 |
> being one, but there are others.) |
30 |
> - This is then us creating a way for those interested people to |
31 |
> contribute in a streamlined fashion. |
32 |
> |
33 |
|
34 |
That's the gist. For the 3rd point I have another POV that I opened up |
35 |
in my previous paragraph a bit: Getting updates to users faster, and |
36 |
providing new exciting packages in moderate. I'm not sure if I remember |
37 |
correctly, but I think discord-bin for example is a package where when |
38 |
upstream releases a new version, they remove the old one from their host |
39 |
and it's also mirror-restricted so for us the package becomes |
40 |
immediately broken when updated. The maintainer immediately makes a PR |
41 |
to update it, but it takes us multiple days to review it while no one |
42 |
can install the latest version from portage. With a simple |
43 |
mirror-restricted package like this, we could cut the wait-time. |
44 |
|
45 |
-- juippis |