Bagas SanjayaHi,
So I'm interested in GPG-sign my commits (that is, ``git commit -S``) for l10n pull request (which I should submit in this cycle). Is it OK to do that? Drawbacks?
Thanks.
-- An old man doll... just what I always wanted! - Clara
threads / discuss / 62665
OK to submit l10n PR with signed commits?
Subject: OK to submit l10n PR with signed commits?
Bagas Sanjaya
Junio C HamanoRe: OK to submit l10n PR with signed commits?
Bagas Sanjaya <bagasdotme@gmail.com> writes:
> So I'm interested in GPG-sign my commits (that is, ``git commit -S``) for l10n > pull request (which I should submit in this cycle). Is it OK to do that? > Drawbacks?
Instead of talking first about drawbacks, we should consider the upsides. Why would we even want to see your GPG signature, when most of us do not even have your GPG public key in our keychains?
What are we trying to achieve by doing this?
Bagas SanjayaRe: OK to submit l10n PR with signed commits?
On Wed, Dec 18, 2024 at 06:49:39AM -0800, Junio C Hamano wrote:
Show 11 quoted lines
> Bagas Sanjaya <bagasdotme@gmail.com> writes: > > > So I'm interested in GPG-sign my commits (that is, ``git commit -S``) for l10n > > pull request (which I should submit in this cycle). Is it OK to do that? > > Drawbacks? > > Instead of talking first about drawbacks, we should consider the > upsides. Why would we even want to see your GPG signature, when > most of us do not even have your GPG public key in our keychains? > > What are we trying to achieve by doing this?
Just to ensure that PR commits are really from the respective authors.
-- An old man doll... just what I always wanted! - Clara
Junio C HamanoRe: OK to submit l10n PR with signed commits?
Bagas Sanjaya <bagasdotme@gmail.com> writes:
Show 14 quoted lines
> On Wed, Dec 18, 2024 at 06:49:39AM -0800, Junio C Hamano wrote: >> Bagas Sanjaya <bagasdotme@gmail.com> writes: >> >> > So I'm interested in GPG-sign my commits (that is, ``git commit -S``) for l10n >> > pull request (which I should submit in this cycle). Is it OK to do that? >> > Drawbacks? >> >> Instead of talking first about drawbacks, we should consider the >> upsides. Why would we even want to see your GPG signature, when >> most of us do not even have your GPG public key in our keychains? >> >> What are we trying to achieve by doing this? > > Just to ensure that PR commits are really from the respective authors.
Yeah, but my point was that it would not ensure, because practically nobody has ways to validate the signature was created with your private key, and public keyservers have been tainted long time ago with fake keys with the same fingerprint, so would not work as a good way to obtain your public key and be sure it is yours.
If this were "because we would want to eat our own dogfood", and if we find bugs in our code when different person sign their commit with their own signature scheme (i.e. you may sign yours with your GPG key, somebody else may use their SSH key, and yet other people use their X.509 certs, it might give us valuable insights, but the resulting history may be irrevocably tainted if the bug is on the signing side (if the bug is on the verification side, that is OK).
Thanks.
Bagas SanjayaRe: OK to submit l10n PR with signed commits?
On Wed, Dec 18, 2024 at 10:02:34PM -0800, Junio C Hamano wrote:
Show 32 quoted lines
> Bagas Sanjaya <bagasdotme@gmail.com> writes: > > > On Wed, Dec 18, 2024 at 06:49:39AM -0800, Junio C Hamano wrote: > >> Bagas Sanjaya <bagasdotme@gmail.com> writes: > >> > >> > So I'm interested in GPG-sign my commits (that is, ``git commit -S``) for l10n > >> > pull request (which I should submit in this cycle). Is it OK to do that? > >> > Drawbacks? > >> > >> Instead of talking first about drawbacks, we should consider the > >> upsides. Why would we even want to see your GPG signature, when > >> most of us do not even have your GPG public key in our keychains? > >> > >> What are we trying to achieve by doing this? > > > > Just to ensure that PR commits are really from the respective authors. > > Yeah, but my point was that it would not ensure, because practically > nobody has ways to validate the signature was created with your > private key, and public keyservers have been tainted long time ago > with fake keys with the same fingerprint, so would not work as a > good way to obtain your public key and be sure it is yours. > > If this were "because we would want to eat our own dogfood", and if > we find bugs in our code when different person sign their commit > with their own signature scheme (i.e. you may sign yours with your > GPG key, somebody else may use their SSH key, and yet other people > use their X.509 certs, it might give us valuable insights, but the > resulting history may be irrevocably tainted if the bug is on the > signing side (if the bug is on the verification side, that is OK). > > Thanks.
OK, thanks! I will stick to unsigned commits then.
-- An old man doll... just what I always wanted! - Clara
Junio C HamanoRe: OK to submit l10n PR with signed commits?
Junio C Hamano <gitster@pobox.com> writes:
Show 13 quoted lines
>>> Instead of talking first about drawbacks, we should consider the >>> upsides. Why would we even want to see your GPG signature, when >>> most of us do not even have your GPG public key in our keychains? >>> >>> What are we trying to achieve by doing this? >> >> Just to ensure that PR commits are really from the respective authors. > > Yeah, but my point was that it would not ensure, because practically > nobody has ways to validate the signature was created with your > private key, and public keyservers have been tainted long time ago > with fake keys with the same fingerprint, so would not work as a > good way to obtain your public key and be sure it is yours.
I think I should rethink this.
Even though I think it is fair to say that more than 99% of people won't have your public key and even if somebody gave them saying "this is Bagas' key", they do not have a way to independently verify it is truly your key (and I think the same thing can be said of my key). But in today's world, there are a few places that it does not matter all that much that you and I do not have each others' keys: hosting sites.
I think both GitHub and GitLab lets you register your public key, so when they are about to show a commit (or a tag for that matter), they can
- notice it is signed; - look up the author/tagger/committer ident of the Git object; - look up the ident in their user database; - find the key(s) of that user account; and - verify the signature using the key(s).
and display the user account that the Git object is signed by a key registered to it.
Now there may be ways to contaminate hosting sites with fake keys that have the same fingerprints as the real ones registered to fake user accounts, and that may render such a feature at the hosting sites less useful. I haven't thought through the security implications.
Of course, $CORP or other organizations can have their members register their public keys and do pretty much the same thing within their closed world. Safeguarding the public key database is their problem so I won't be worried about, unlike hosting sites where practically anybody and their dogs can create accounts ;-).
Thanks.