Features Download
From: Alexander Belchenko <bialix <at> ukr.net>
Subject: Re: Bzr development stopped
Newsgroups: gmane.comp.version-control.bazaar-ng.general
Date: Sunday 11th November 2012 07:45:39 UTC (over 5 years ago)
I'm sorry for being late for 2 months and only now trying to say
something. I'm sorry for being grumpy or saying something unpleasant.

When I saw the subject of this thread in the mailing list I was very
sad, because it was just a confirmation of what I knew before. So I
was unable to force myself to read all this thread, just to avoid even
more sadness. I read it now and while I see positive tone in
discussion, but the outcome is still rather pessimistic.

I never was Canonical employee, but I worked on bzr some time as some
of us knows. So please forgive me my open-hearted mail, but what I saw
in 2007, 2008 and so on - I didn't really like what was going on. In
short: there were a lot of promises what bzr can do, but bzr never
reached those goals. I stopped advocate bzr in 2008 or maybe 2009.
Because I didn't have a real arguments.


There is still nested trees implemented but unmerged lying on launchpad.

There is still no proper solution for line-endings conversions, or
keywords. I've tried to push the ball in 2008 but it was my own the
biggest fiasco.

There is no [simple] setup for local private bzr server where I can
put my working projects, and bzr-access script in contrib/ is rather
too basic and limited. Today on my last job where I've been using bzr
since 2006 I run dead simple `bzr serve --allow-writes` just because I
am either too stupid to setup the proper access control or too lazy to
dive in and experiment when I know this needed only for couple of

There is no third-party site that allows me for small fee to host my
private projects, and hosting private projects on LP did never seem
like a real answer, mostly of the lack of visibility of this feature
and hence availability for people. I was under impression (possibly
wrong) it was discouraged by Canonical itself.

I think the crucial point was in 2009 when it was obvious for
Canonical management that bzr is unable to beat hg/git, so most of the
efforts was put into UDD. Since then I saw a real change in the bzr

What I see today? git is everywhere and that means that I have to
learn it. Where is bzr? On Launchpad [only].

For me bzr is not going to really compete in the near(?) future. I did
feel this for last 3 or 4 years. And that made me very sad all this
years. I've spent too many years working on bzr and other related
stuff in my spare time. But last year was stagnant even for me. Today
I can't find the reason to continue work even on qbzr. I feel like I
stay in the village while all citizens left it, although some
strangers sometimes arrived and moved on quickly.

Most of bzr hackers I used to know since 2005-2007 are not in this
boat anymore. Martin, Andrew, Robert left Canonical. For their own
reasons, but. Ian passed away. Guys who worked on qbzr/explorer stuff
  with me are not here anymore. No wonder I feel sadness.

About Contributor Agreement.
As a developer I worried about this contributor agreement before sign
it. Just because I don't understand this legal stuff. But then just
sign it and moved on. Just add to the statistics.

And yes, the hardest part is writing proper tests. I know it's very
important, and that taught me a lot. But it was always hard. And also
the complexity of the project itself. There is a lot of good features
in the code, but overall there is too much code to work with. And
sometimes I think that strong focus on backward complexity is also
makes bzr hacking very hard. Backward compatibility is very good for
users (and I'm user too), but it makes developer's life much more harder.

I know the tests are important for the project itself. But for happy
users that is not the most important thing. (Sorry Vincent)
CD: 3ms