12 May 2008

The Art of the Open Source Release

Mark Shuttleworth writes about another advantage that free software sets enjoys over monolithic, proprietary code collections:

An update on the long term plans for Ubuntu release management. 8.04 LTS represented a very significant step forward in our release management thinking. To the best of my knowledge there has never been an “enterprise platform” release delivered exactly on schedule, to the day, in any proprietary or Linux OS. Not only did it prove that we could execute an LTS release in the standard 6-month timeframe, but it showed that we could commit to such an LTS the cycle beforehand. Kudos to the technical decision-makers, the release managers, and the whole community who aligned our efforts with that goal.

As a result, we can commit that the next LTS release of Ubuntu will be 10.04 LTS, in April 2010.

This represents one of the most extraordinary, and to me somewhat unexpected, benefits of free software to those who deploy it. Most people would assume that precise release management would depend on having total control of all the moving parts - and hence only be possible in a proprietary setting. Microsoft writes (almost) every line of code in Windows, so you would think they would be able to set, and hit, a precise target date for delivery. But in fact the reverse is true - free software distributions or OSV’s can provide much better assurances with regard to delivery dates than proprietary OSV’s, because we can focus on the critical role of component selection, integration, testing, patch management and distribution rather than the pieces which upstream projects are better able to handle - core component feature development. This is in my mind a very compelling reason for distributions to focus on distribution - that’s the one thing they do which the upstreams don’t, so they need to invest heavily in that in order to serve as the most efficient conduit of upstream’s work.


Anonymous said...

All very well, but it sounds like Shuttleworth is trying to justify the fact that Canonical does very little upstream development, as compared with the other distributions.

Glyn Moody said...

Well, I think his point is that he's engaged in meta-development: that is, at the distro level.

Anonymous said...

I don't think comparing Ubuntu & Windows development really works too well. Microsoft put the features they are going to implement on the table and then try to get something out that matches the feature set.

Ubuntu is a collection of pieces cobled together from other places and integrated. If a particular feature is in the pieces as they are assembled great, if not no big deal they'll get it next release.

I'm not saying one is better than the other, it just isn't the same thing so it seems pointless comparing the relative timeliness of delivery.

Glyn Moody said...

You're right, but I think this is what Shuttleworth is saying: that Microsoft is constantly fighting to keep up with its PR, whereas Ubuntu just takes what's available and uses that. What's interesting is his idea of coordinating distros, so that the pressure moves on to the upstream developers: if you want to get into the next mega-release, you'd bet get coding....