Re:Git 3.0's upcoming SHA-256 default will be a costly mistake
by lxgr on Friday October 02, 2026 @04:31AM
(#49931193)
> Git hashes are not supposed to be a security mechanism.
They absolutely are, both when you're using signed commits (i.e. GPG, SSH, and S/MIME) or just identifying a given commit/repository state by hash via a secure channel and then serving object data over untrusted transports.
It's entirely possible that you don't use either, but that's certainly not true for everybody.
Calling anyone using either to be "doing it wrong" is borderline gaslighting: Git used to have these security guarantees, and just because they're now broken doesn't mean they were never there in the first place, or that it was stupid to rely on them.
They absolutely are, both when you're using signed commits (i.e. GPG, SSH, and S/MIME) or just identifying a given commit/repository state by hash via a secure channel and then serving object data over untrusted transports.
It's entirely possible that you don't use either, but that's certainly not true for everybody.
Calling anyone using either to be "doing it wrong" is borderline gaslighting: Git used to have these security guarantees, and just because they're now broken doesn't mean they were never there in the first place, or that it was stupid to rely on them.