Re: [PATCH] daemon: return "access denied" if a service is not allowed

From: Jeff King <peff@peff.net>
Date: Oct 14, 2011, 20:34 UTC
Message-ID: <20111014203438.GA15643@sigill.intra.peff.net>
In-Reply-To: <7v7h47z5i0.fsf@alter.siamese.dyndns.org>

On Fri, Oct 14, 2011 at 01:24:07PM -0700, Junio C Hamano wrote:

Show 10 quoted lines

> > Git is foremost an open system, and our defaults should
> > reflect that.
> [...]
> 
> I think the logic in the last paragraph is flawed.
> 
> There is a difference between Git being an open system, and installations
> and users of Git being primarily people who work on open projects.
> 
> Even though personally I wish there weren't.

I think it is not the logic that is flawed, but the communication. What I meant was that git was originally designed to support open projects (like the kernel), and they are our primary target.

Ingo said something similar here:

  http://article.gmane.org/gmane.linux.kernel/1202320

Still, primary target and primary user are not necessarily the same thing. And a minor convenience for one audience that introduces a security problem for another audience may not be a good tradeoff, no matter who the audiences are.

I didn't really expect you to take my second patch. We tend to be a bit more conservative than that around here.

Show 6 quoted lines

> > But since it is a potential security issue, it does seem kind of mean to
> > closed sites to just flip the switch on them.
> 
> It would have been a better split to have the 1/2 patch to support both
> informative and uninformative errors, with the default to say "access
> denied", and 2/2 to flip the default to be more open.

Isn't that what I did? It was what I meant to do, anyway...

Or did you mean the options would have been better worded as:

  --errors={terse,informative}

or something similar?

-Peff