1 |
On Mon, 2005-08-15 at 13:54 -0400, Jeffrey Forman wrote: |
2 |
> Oh $diety, please let that be how it's done in the future. If I am still |
3 |
> around and haven't lost my mind, this is how it should be done. A clean |
4 |
> handoff from releng -> infra is what is needed so that we arent in any |
5 |
> confusion about who's doing what... |
6 |
|
7 |
Absolutely. |
8 |
|
9 |
Also, I think we need to impose a policy on updates to things |
10 |
in /releases. Over the past two releases, we've had things that got |
11 |
updated on /releases without doing any sort of renaming of the files. I |
12 |
think this causes bugs and confusion for our users. Instead, we should |
13 |
enact and enforce a policy that no two different files in /releases |
14 |
should ever have the same name. It'll cause a bit more bandwidth usage |
15 |
for the mirrors, as they won't be able to simply rsync differences in |
16 |
the file, but it will also reduce headaches anyone might have. |
17 |
|
18 |
Example: stage3-athlon-xp-2005.1.tar.bz2 would become |
19 |
stage3-athlon-xp-2005.1-r1.tar.bz2 |
20 |
|
21 |
This would also help enforce the releng->infra hand-off, as we would |
22 |
simply give you the new file, and you guys would push it out, along with |
23 |
updating the bouncer. At the same time, we would *require* that any |
24 |
such changes be given to the GDP for them to update their information. |
25 |
|
26 |
Now, should infra (well, Jeff) take the point in coordinating changes, |
27 |
as in commits of updates, with the GDP, as it will be determined by the |
28 |
time things are pushed to the mirrors, or should that remain a task for |
29 |
the releng leads? |
30 |
|
31 |
-- |
32 |
Chris Gianelloni |
33 |
Release Engineering - Strategic Lead/QA Manager |
34 |
Games - Developer |
35 |
Gentoo Linux |