[rfc] Readonly/mirror branches

[Martin Pool]

On 09/05/2006, at 12:55 PM, Robert Collins wrote:

> On Tue, 2006-05-09 at 12:31 +1000, Martin Pool wrote:
>> On 09/05/2006, at 8:58 AM, Robert Collins wrote:
>>
>>> I'd like to just completely remove the 'push to the branch we  
>>> bind to'
>>> logic.
>>>
>>> Given that we have bzr push and bzr commit both able to push into  
>>> the
>>> master, it seems like bind is serving two masters : it is  
>>> establishing
>>> the link, and it is performing update and push as well. I think bind
>>> should *just* establish the link - like unbind *just* breaks it.
>>
>> That's probably simpler, but we should make it clear what happened.
>> If there are new revisions there we should say something like
>>
>>      Bound to http://fofofof/
>>      There are 12 new revision in there; use bzr update to bring  
>> them in
>>
>> If they've diverged we should give an error as at present.
>>
>> This also relates to the question of just what should be shown by  
>> log.
>
> I think if they are diverged, the local commits should be treated as
> local commits, and no error should occur.

OK - no error, but there should be a notification of what's happened.

> Only when there is no common ancestor other than NULL_REVISION  
> should we
> generate an error.

OK.

-- 
Martin