metronome mail for Bazaar 2.0 (or 1.18) - 6 August

[Martin Pool]

2009/7/24 Jean-Francois Roy <bahamut at macstorm.org>:
> For what it's worth, I'll throw a +1 on that. Getting the core
> functionality, the "basic workflow", 100% kink-free is absolutely critical
> to adoption. I've witnessed it personally: if anything goes wrong within the
> first few days of usage, there's a good chance the person will drop Bazaar.
> This also includes a serious review of things like default behaviors and
> upgrading people to 2a (what good will it do if people don't upgrade?).
> Things like "Oh, you have to create a repository first to get good local
> branching performance, oh and you need to specify --format=1.14-rich-root
> (please don't ask what rich root is), oh and you may want --no-tree to be
> like Git, and oh you need to get this plug-in, which has this dependency and
> requires this number of settings to make this work great" are a burden on
> adoption.

So thanks a lot to you and Andrew for the interesting feedback.  I
thought a lot about it over last weekend.

Perhaps the most provocative point is that 2.0 is not going to have
enough news or buzz.  I think if we just keep releasing what's in
trunk every month, any single change is always going to be small.
That's good for testing but bad for making a splash.  Therefore, if we
have a cycle of releases at 6 month intervals, each one should have
one or more large features that are new to people running the stable
releases, and it makes it a bit more interesting.

There are a few things here that come under the general heading of
polish and user experience, but that need to be done differently.  The
upgrade experience to 2a is important; Ian put up at least one patch
for that and we will do more.  Feedback on cases where upgrade doesn't
work well, or testing of it, is very welcome and can be targeted to
2.0.

There are also many polish-type bugs that can be fixed individually.
For instance, that you get a traceback when some unexpected network
errors like host unreachable occur - you can live with this but it's
not a good look and we should give a cleaner message.  There was some
discussion a while ago about why bugs tagged 'easy' are not getting
done faster, and we should perhaps do a 100-papercuts type project to
close more of them off.  I don't think it's reasonable to block 2.0 on
this type of bug, which has a nearly inexhaustible supply, or to have
a long discussion about their relative priority.  If people put up
clean patches for them they can certainly be merged.

I would agree with AfC that there are hundreds of such bugs.  I am
sure we would not close them all if we slipped by a month or even two
months - we would probably make a pretty big dent, but people would
still sometimes hit them.  So I think it's something that has to be an
ongoing effort and gradually getting more efficient and faster, rather
than something to do just before release.

Then there are larger questions like having separate repository
directories and whether you should use a separate repository with
--no-trees.  I agree they need some love.  I believe while we can make
some piecemeal improvements, properly solving it really depends on
having a clear description of how the system ought to appear to the
user and aligning everything with that.  That's really important and
I'm looking forward to do doing it, but realistically it's going to be
multiple months of development with some instability while it's
underway.   I'd like to set a stable checkpoint first where people can
get the accumulated improvements, to free us up to try this with the
users who want to follow it.

> May also want to consider the "batteries included" idea and make sure Bazaar
> comes well-equipped in terms of marketing points (yes, really). "Bazaar 2.0
> has (yet another) better format!" isn't too "buzz-worthy".
>
> Just my thoughts and humble opinion. I totally respect the huge amount of
> work and functionality that has gone into bzr since the 0.92 days. It's a
> fantastic tool (even more so with Launchpad, which should be advertised even
> more!).

I'd like if you would expand on how you think we should include
batteries, present marketing points, and promote Launchpad - but maybe
in one or more separate threads from the metronome one.

-- 
Martin <http://launchpad.net/~mbp/>