@fribbledom "maintaining our
own git infrastructure is an unnecessary security risk"
If they can't do it, then who can?...
@fribbledom I mean as long as I dont need a github account to participate thats ok.
I'm honestly concerned about that monoculture on github.
For example if some country now decides to block US services, you will not be able to contribute to PHP anymore, merely because they depend only on github instead of an independent, self hosted service.
@sexybiggetje With mercurial, you can create a "bundle" which consists of just chosen commits. On the receiving end, they can pull those commits in as if the bundle were a full clone. Does git do something like that?
@sexybiggetje Cherry picking exports a file? I thought it was just for destroying your compatibility with upstream. Maybe I'm thinking of rebasing.
Do you mean creating a patch file of specific commits? Those could just be done by Exporting git diff for specific commits to a file
@sexybiggetje I guess so, but the `hg bundle` command creates a binary file, not text you could potentially edit. And is easy to pull from. But also, pulling in mercurial doesn't update the bookmark (branch), or working directory. So in hg workflow, you often compare incoming commits after pulling, not before. Which means git users probably don't do things like this, huh?
I think what’s being described here is most like git format-patch
git can also create a patch from a set of commits and email it.
Lots of options for creating patches and then applying them.
I understand. That's not possible. But I don't see much problems in that when offsetted to the point of maintenance/upkeep they have to do.
You provide a valid problem, but the same could happen to for instance hosted gitlab. Hosting over multiple vendors is an admin nightmare as well. So the only way is to self host. Which could ofcourse also be blocked.
Short summary: I don't think there is a perfect solution that satifies all problems.
Horrific. Ideally need to find a way to get commit signing to be required, and enforce specified public key, so at least compromising undetected is much harder.
If git host is pwnd they could still swap out public key, but then interface should show past commits as unverified etc. and that kind of key change should be broadcast. Also can check against public keyserver so could detect suspicious changes, and disallow commits that don't match.
How else do you verify identity?
@fribbledom Putting Microsoft in control will surely fix our security problems...said no intelligent and sane being ever...
@fribbledom The real question is how they managed to tell the difference between a malicious and legitimate commit in php. Presumably their suspicions were raised when the code quality unexpectedly went up.
@fribbledom Terrible! Not only were they hacked, but also their response is to subjugate their development community to Microsoft!
(One wonders if Microsoft wasn't actually behind it all)
I think you're misunderstanding this. It's certainly not normal that someone commits exploits under the project maintainer's name.
Server run by the main developers of the project It is not focused on any particular niche interest - everyone is welcome as long as you follow our code of conduct!