On Sun, May 31, 2009 at 10:56 AM, ron minnich <rminnich@gmail.com> wrote:
On Sat, May 30, 2009 at 9:01 PM, Peter Stuge <peter@stuge.se> wrote:

> I would like to request a better patch management system than this
> mailing list. The fact the above "ping" is now a part of our
> development process is a very strong indication that things are not
> functioning very well.

it's a hard problem. I'm on several projects. They are all non-ideal
in some way. Linux sucks in patches at the rate of 30,000 a year or
so; that's fine performance but some feel (me included) that the
kernel is "de-cohering": it no longer has the small tight feel and
coherence of vision that it might have once had. Plan 9 still has the
same tight feel and coherence but at a cost: important patches seem to
linger on the vine for (i am not kidding here) years .

Coreboot is trickier than a kernel, as trivial errors can lead to
systems that can not be recovered. I especially avoid acking flashrom
patches because I can't test most of them. Others I know don't like to
NAK, but they're not comfortable with an ACK either; they don't like
the code but they don't want to hold up progress.

All in all, I think the process works. Yes, it is not ideal.  Yes, it
could be better, but so could everything.

ron

Indeed, changing the whole process might not be worth the effort considering the trade-offs. There are some tools that can make the current process a lot less painful, however, specifically web-based code review dashboards. Review Board from the folks at VMWare looks really good, and Rietveld, which is a relatively new open-source fork of Google's Mondrian code review tool, is quite helpful as well though it seems behind RB at the moment.

--
David Hendricks (dhendrix)
Systems Software Engineer, Google Inc.