2009/12/18 John Arbash Meinel <j...@arbash-meinel.com>: > James Westby wrote: >> On Thu, 17 Dec 2009 14:19:32 +1100, Martin Pool <m...@canonical.com> wrote: >>> My hypothesis is that we (Canonical's Bazaar team) will get to grips >>> with UDD better if there is a tighter medium-term focus. >>> >>> The key question is whether vcs-imports is that thing, or not. >>> >>> If we do make imports that one important thing then I'll be asking >>> people not to work on looms, UDD merge support, or nested trees until >>> they're done. >> >> What would you consider "done." Do you think that with 4 months effort >> you could get to 0 outstanding failures, or would there be some other >> stop point? > > I would guess that needs to be evaluated on the fly. If we are getting > to the point of diminishing returns, then it is reasonable to > re-evaluate what our priority should be.
+1 The input data is messy enough that we may never get it to zero failures, so maybe "done" is a poor word compared to "enough": defined as the point where other things are becoming clearly more important, or the remaining bugs have a poor cost:benefit. We may actually already be at that point now. Since it's a long tail problem, we don't want to set up our goal as having zero failures or doing vcs-imports as an end in itself, but rather to relate it to larger needs. Thus this thread. There's some up front cost to getting into the analysis of import failures, so if we do focus on that we would want to persist with it for a few months. It's hard to estimate until we get further into it but it's possible that four months would get many of them cleared. It's a fairly large commitment and opportunity cost. I do think Elliot has a good point though, that fixing imports is not going to really teach us important things about the structure of the problem. If we had UDD or build-from-branch totally working on some packages, but people were blocked from using it widely because of failing imports, then it would be clearly the right time to do it. -- Martin <http://launchpad.net/~mbp/> -- ubuntu-distributed-devel mailing list ubuntu-distributed-devel@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-distributed-devel