[Kde-scm-interest] Access control

[Ian Monroe]

2009/8/12 Jeff Mitchell <mitchell at kde.org>:
> Johan,
>
> Had a discussion with some KDE people and had a few thoughts regarding
> possible access control/hook solutions:
>
> 1) You don't want to allow arbitrary shell scripts -- but might there be
> a way to allow sanitized text with specific wording, or entered into a
> specific UI in the webapp, such that pre-commit and post-commit hooks
> could be used? A specific issue we have is a repository that is
> currently svn-externed into kdelibs but that we cannot allow arbitrary
> write access to for any kdelib developer. That way we could possibly
> have specific data values that could be converted on the back-end to a
> pre-commit hook so that we can limit write access to a particular path.
> We don't really want to put it in another repo and use a git submodule
> because that creates a burden on everyone cloning the repo.
>
> 2) I was thinking about how you were talking about sending out commit
> information via HTTP. If we set up pub/pri key pairs, could something
> like the following be done: in a pre-commit hook, you send out
> information via HTTP, with a secret encrypted with a public key and a
> callback URL.

If you sent the information via HTTPS then you would have verifaction
that you're talking to the correct server for free.

> The server you send it to runs a script to validate things
> -- for instance, person X has authorization to commit to path Y -- and
> sends back either an approve or reject message, along with the decrypted
> contents of the secret for verification. Then the commit is allowed or
> denied based on this. This would allow us to do arbitrary processing
> with the commit data and not have to put that processing on your machines.
>
> Thanks,
> Jeff