On 7/4/06, John Levon <[EMAIL PROTECTED]> wrote:
On Tue, Jul 04, 2006 at 01:39:32PM +0200, Asger Ottar Alstrup wrote: > John Levon wrote: > > No review is done > > This is better than code does not get committed, given that the > developer is competent. There really is no developer I've ever met who doesn't need review, and I work with some very very smart developers...
If this is a pissing contest, I know many people whom I would trust to commit without review. I believe any responsible, competent developer can be given that priviledge. My experience is that developers become *more* responsible when they get this priviledge, because they know they have to act safely in an environment without a safety net. Most people know when they are on shaky grounds, and then ask for a review themselves. But all people make mistakes. Reviewers more so than developers. The task is to design the system such that the error are prevented at reasonable cost, including errors in the review system. There is an extensive body of research on this. The aviation industry has been working with these issues for years. They have learned the hard way that reviews on their own do not cut it - the planes kept crashing down. It takes a change of attitude. Part of this is to demonstrate again and again that you are competent and act safely, and conversely the system also needs to allow people to act on their own. Regards, Asger
