Re: We Want Quality

On Thu, 29 Jun 2000, Andrei Zmievski wrote: > On Thu, 29 Jun 2000, Zeev Suraski wrote: > > Well, I do consider this not only a big problem, but something that's > > simply not doable. Moreover, I don't even consider this a positive thing. > > Releasing a patch level a day after the release is, in my opinion, much > > better than freezing development at different levels for a couple of > > weeks. Considering there were 2 such fuckups (a major one with gpc back > > then, and a minor one with the Windows CGI now) on more than 20 releases, > > I think our process is more than reasonable as a baseline. > > I'd once again disagree. I don't know if you've ever gone through a > commercial alpha/beta program, but I have - in fact, I'm going through > one right now for a big company. The rule is that after a certain date, > no features get added and no new stuff gets committed, except for fixing > major crash bugs and such, and even then, some of them are put off until > the next incremental release. Granted, our codebase is much smaller, so > we like to think that we know it pretty well and by implementing a minor > feature or a bug fix, we will not destabilize the tree. However, as you > pointed out, it did happen a few times already, and it will happen > again. I see nothing wrong with being a little more professional and > disciplined as far as our release process is concerned. Don't twist my words please; It happened twice out of a couple of dozen releases, one was a fairly major showstopper, the other was not. Overall, I'd say that it's working pretty well, and it doesn't fall below commercial standards. I also don't think our codebase is small, and is in the commercial neighbourhood easily. Not every project is Windows 2000 with 30 million lines of code. > > Considering nobody cared about the 4.0.0 release schedule, I find it > > surprising you say we can adhere to a schedule. We pretty much proved > > that we can't. > > Perhaps, schedule was a bad choice of a word - I meant that it wouldn't > be too hard to close down CVS for 5 days or so and make sure that > we do not screw up the RC accidentally. I disagree. Cooperation from the developers with the current release scheme would do the trick. Our main issue is developers who don't fix their bugs (see the bugs database), and not last minute changes. You can't blame them either, we're on a free-will based project, after all. The amount of damage made by last minute changes in the last 3 years was negligable, even though we should try to reduce it even more, but not by draconian rules. Zeev -- Zeev Suraski <zeev@zend.com> http://www.zend.com/