1 |
On Tue, 16 Oct 2012 22:54:04 +0000 |
2 |
"Robin H. Johnson" <robbat2@g.o> wrote: |
3 |
|
4 |
> Previously, the PORTAGE_GPG_KEY variable has allowed ANY argument, and |
5 |
> passed it to GPG, letting GPG use that. This was intended to explicitly |
6 |
> be a unique identifier for a key (or subkey). |
7 |
> |
8 |
> However, it seems that there are signed commits with other values in the |
9 |
> variable, and instead of something nice like: |
10 |
> (Portage version: 2.2.0_alpha138/cvs/Linux x86_64, signed Manifest commit |
11 |
> with key 0x586A3B1F) |
12 |
> We have commits with: |
13 |
> (Portage version: 2.2.0_alpha138/cvs/Linux x86_64, signed Manifest commit |
14 |
> with key emailaddress) |
15 |
> |
16 |
> This makes validation harder, as we need to extract the identity of the |
17 |
> key from the Manifest before we can proceed. Additionally, if a |
18 |
> developer has multiple keys, possibly over time, we cannot use this |
19 |
> string to identify what key was used easily. |
20 |
> |
21 |
> As such, we've decided to make the PORTAGE_GPG_KEY strictly enforce what |
22 |
> was originally intended. |
23 |
> |
24 |
> - You must specify a key or subkey exactly. |
25 |
> - The leading "0x" is optional. |
26 |
> - If you want to use a subkey, per the PGP specifications, you must |
27 |
> suffix your keyid with "!". |
28 |
> - Your keyid is exactly: 8, 16, 24, 32 xor 40 hexdigits long. |
29 |
|
30 |
Isn't that fixing the issue from the wrong end? |
31 |
|
32 |
I agree that the keyids in commit messages should follow some kind |
33 |
of spec. But I rather think that portage should be modified to convert |
34 |
any supported argument to follow that spec rather than the spec being |
35 |
forced into the configuration file. |
36 |
|
37 |
Also, will that matter anymore after the git conversion? |
38 |
|
39 |
-- |
40 |
Best regards, |
41 |
Michał Górny |