On Sat, Jun 14, 2008 at 11:33:09AM -0700, Andrew Lentvorski wrote:
The second article finally articulates the bad "gut feel" that I have
always had about Linux development. It was an "anti-BitKeeper" article
that I somehow missed on the first round. The argument, rather than being
concerned by the license, revolves around BitKeeper encouraging a specific
style of development (just like Linux) that may not be the best option.
http://web.mit.edu/~ghudson/thoughts/bitkeeper.whynot
Good article, although I feel it misses the point.
I see two issues here:
- Reviews. The problem with the multiple-committers model is that code
usually ends up not getting reviewed by anyone. The Linux model does
make this happen, admittedly with the problems associated with
filtering the reviews through a small number of people.
- Hierarchy. This has both advantages and disadvantages. When the
maintainers have time and are indeed experts, it greatly increases the
quality of the code. I've seen this work on other projects as well.
It has problems when the maintainers get overwhelmed, or have personal
issues that get in the way.
I encourage anyone interested in methodology to look at Aegis. I feel the
tool itself is kind of clunky and difficult to use, but I have found the
methodology behind it to be quite good.
We are using something like this model, using a git-type development to
enforce it. The maintainers aren't the ones who review the code, but
simply make sure that the proper people have reviewed the code. I may
figure out how to automate that at some point.
Having the more experienced engineers have to review things before they are
allowed to be checked in is important.
David
--
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
