Problems with hg push over SSH
Dustin Sallings
dustin at spy.net
Wed Aug 29 12:05:02 CDT 2007
On Aug 29, 2007, at 3:02 , Paul Sargent wrote:
> [a whole lot of scary chaos that boils down to this]
>
> If it's *your* project, you can tell people where they can
> get your
> code. If it's someone else's project, they might want a bundle or
> patchbomb or something. If you could host your own branch on a web
> server, it's often better.
>
> I'm fast coming to the conclusion that the randomly connected
> distributed model just doesn't scale, but because the tool is
> distributed you have the choice of topology that works best for
> you. Maybe that'll be a tree, like Linus uses for the Linux Kernel.
> Maybe (shock, horror) it'll be a centralised repository. It'll
> probably end up being some hybrid. The main thing is to actually
> think about it, because if you let it grow naturally it'll be chaos.
Sorry, I was home sick yesterday and used that as an excuse to
ramble on with anecdotes.
It really just comes down to your project. The distributed model
scales both up and down better than the centralized model.
In one of my anecdotes, I mentioned the ``couple of guys sitting in
a room'' model. In that case, setting up a centralized repository
with accounts and pushes and all that would've been a distraction and
not useful to our cause. Instead, each of us ran ``hg serve'' and
pulled from each other.
In the *really* large team models, having a single centralized
repository is as equally counter-productive as having it be fully
distributed. In this case, you may have a central ``release'' tree
that effectively nobody writes to, but a release engineer pulls from
a ring of component-specific release trees that component owners
update by pulling from integration trees that component engineers
push to.
Given that, you may be able to come up with a formula like, ``you
must have one centralized tree for every 3-16 developers,'' but I
don't believe you will find a rule to which someone won't take
exception.
--
Dustin Sallings
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://selenic.com/pipermail/mercurial/attachments/20070829/cdc4046f/attachment.htm
More information about the Mercurial
mailing list