IRC channel logs

2021-07-08.log

back to list of logs

***rt is now known as robin
<Hagfish> https://security.googleblog.com/2021/07/measuring-security-risks-in-open-source.html
<Hagfish>interesting idea
<Hagfish>let's hope it incentivises the right behaviours, over all
<stikonas>well, but you can always game that score a bit if you are malicious. E.g. for signed releases thing it says "The check does not verify the signatures." So even completely broken signature would pass that check
<civodul>yeah, it reminds me of the "Verified" badge on GitHub
<hendursaga>civodul: did they finally make a way where you could verify that badge offline by yourself with PGP??
<stikonas>hendursaga: well, they don't have to make any way, you just do that with git/gpg
<stikonas>but as usual with gpg, you want your key to be in the strong set to be able to verify anything
<hendursaga>stikonas: I meant, when the maintainer doesn't provide signed hashes in the releases page or elsewhere
<stikonas>in that case how can you check...
<stikonas>you need at least signed git tags
<hendursaga>because isn't the verified badge only when they sign commits or releases or something?
<stikonas>or in some cases (like package manager repos in e.g. guix or gentoo) you need each commit to be signed
<stikonas>it can be either commits or tags
<hendursaga>hmmm, so when that badge is shown, the signed commit/tag is somewhere in the Git repo that you could then check?
<stikonas>yes, with git log --show-signatures
<stikonas>--show-signature, apparently it is singular
<hendursaga>oh, TIL - I'm only used to signing commits and sending them through email (I used to do some Guix development)
<stikonas>I don't think commit signature survives email...
<stikonas>since commit is actually made by somebody else
<stikonas>so has to be modified
<stikonas>git has both author and committer fields
<hendursaga>no, like, creating a patch file and signing it then sending both..?
<stikonas>I think that's different from git signatures
<hendursaga>yes, that's what I was trying to convey
<stikonas>it's the same idea and algorithm...
<hendursaga>I'm currently rather excited about Pijul, eventually they'll have multiple ways to sign/verify patches and all that
<civodul>hendursaga: re the OpenPGP badge on GitHub, i don't know
<civodul>but i think it's pointless without a mechanism similar to what Guix has
<civodul> https://guix.gnu.org/en/blog/2020/securing-updates/