S.Ramaswamy wrote:
Interestingly with a distributed model, the notion of commit access may not
be that important at all. Nobody commits directly to the Linus git
repository of the kernel except himself yet we have that model scaling
incredibly well. Just a side note.
This line of reasoning is slightly bizarre.
official release is not just a matter of cloning a repository and then
committing changes locally with a distributed version control system.
Regardless of the nature of the version control system there are
always gatekeepers who choose to accept or reject your changes. In the
case of subversion it's the committers and in the case of Linux it's
Linus' gatekeepers like Andrew Morton. I can clone the Linux git tree
and commit locally, but one of the kernel gatekeepers has to accept
it. Otherwise it's not going to get into the official Linux kernel.
I think the distributed version control guys have not been able to
articulate well their case well, and have been throwing out this "You
don't need commit access with DVCS" marketing bit. There actually
might be a few scenarios for which the DVCS approach is suited (like
the Linux kernel), but those have not been explained well so far
AFAIK.
