Re: Github mirror: how does it work?

Le 02/07/2013 19:51, Peter Linss a écrit :
> On Jun 25, 2013, at 12:14 PM, Tab Atkins Jr. wrote:
>> On Tue, Jun 25, 2013 at 11:20 AM, Peter Linss<peter.linss@hp.com>
>> wrote:
>>> I've been exploring what it would take to make the bridge
>>> bi-directional, allowing pushing to either repository. There have
>>> been some attempts which work for simple cases, but there are
>>> gotchas dealing with branches...
>>
>> We shouldn't worry about branches - only the master branch should
>> be canonical.  Ideally, we should reject pushing non-master
>> branches to the repo.
> Unfortunately it's not so simple. I'm also not just talking about
> long-term, named branches, but the local, unnamed branches that are
> just part of the workflow of a dvcs. We can only enforce the end
> state of what someone pushes, if we try to enforce intermediate
> steps, it's way too easy for someone to get their local repo into a
> state where they can never push again. There's nothing preventing
> people from creating and merging branches locally.

I don’t see what your concern is, Peter. We could reject pushing to 
non-default branches on dvcs.w3.org/hg/csswg (and non-master branches on 
GitHub, if they have that feature), that wouldn’t at all prevent people 
from having many branches in their local repositories.

Now I’m not sure we *want* that, but if it helps with hg/git interop I’m 
all for it.

-- 
Simon Sapin