Access control to parts of the repository

[Jesse Glick]

Bryan O'Sullivan wrote:
> The current strategy, conservative as it is, is arguably better from
> an access control point of view: you can't mess in any way with the
> history of a file to which you don't have push access.

But if the doc writer and the programmer are using the same branch in 
the same repository, effectively only one of them can do any work. This 
would seem to make the ACL pretty much unusable, wouldn't it? By 
comparison, in Subversion if you restrict write permissions to subdirs 
it is fine, even though revision numbers are repository-wide, because 
the implicit merge done by 'svn commit' effectively bypasses the ACL.