| 1 |
On Tue, 2020-10-06 at 18:17 +0000, Robin H. Johnson wrote: |
| 2 |
> While I'm absolutely in favour of the overall intent here, I'm not so |
| 3 |
> sure of the design. |
| 4 |
> |
| 5 |
> I'm worried about the proliferation of tiny packages just to convey the |
| 6 |
> keys; and how versioning should work if upstream rotates their keys. |
| 7 |
> I picked this message in the thread to respond to, because it was |
| 8 |
> clearest that this could break when the keys are rotated. The old |
| 9 |
> releases might not be verifiable with the new keys. |
| 10 |
> |
| 11 |
> Additionally: |
| 12 |
> - not all upstream providers ship .asc files of their keys |
| 13 |
> - some upstreams use signed DIGESTS files rather than directly signing |
| 14 |
> the distfiles (esp. where distfiles are larger) |
| 15 |
> |
| 16 |
> Can we instead: |
| 17 |
> Inside the ebuild and/or metadata.xml: convey: |
| 18 |
> 1. URL(s) to fetch keys, incl a keyserver support |
| 19 |
> 2. Full key fingerprint |
| 20 |
|
| 21 |
It won't work inside network-isolated environments. |
| 22 |
|
| 23 |
-- |
| 24 |
Best regards, |
| 25 |
Michał Górny |
Archives