Shuttleworth on Debian

A [great post](http://www.markshuttleworth.com/archives/56) by Mark Shuttleworth, on how he thinks of Debian and it’s relationship with Ubuntu. I wish I was as coherent a writer – he manages to stretch the “altitude” analogy so far it nearly snaps, but not quite :)

Where I think Shuttleworth is perhaps wrong is his belief that the bickering results purely from conflicting goals. Although debian-devel can be pretty shocking, the worst I have ever seen for vicious ad hominem attacks is the various [OpenBSD](http://www.openbsd.org) lists. [Theo de Raadt](http://www.theos.com/) is, unapologetically, a complete bastard on those lists, and will happily berate those who annoy him, and bystanders, pretty nastily.

Theo takes it too far, but he has a point in his unapologeticness, which he has explained sometimes. Quality is vital in an OS distribution, and a lot of the work is very difficult to check. The level of understanding you need to complete a piece of work is immense. Having seen what is required to do this properly, it’s amazing anyone does it for fun. To package a significant piece of software can take weeks or months of research, to completely understand the package. Then you have to understand all of the common use cases, and then you have to understand the whole packaging and distribution system, and the implications of all of the above combined.

There is so much involved, that often there is (after the research) only a single person in the world qualified to check the work – and that’s the person doing it. So, how do you, as an organisation, ensure the quality of the work?

You can have intense vetting procedures, of course, but these only take you so far. Even a quality individual can be careless sometimes. You can perform audits — but you probably aren’t qualified to do them. If you are paying people, or have contracts with them, you can threaten horrendous consequences if they fail to do the work well enough — but these are volunteer organisations.

The only way these organisations can enforce quality is culturally – by having a balance of *cultural* rewards and sanctions that encourage individual quality. At 11.30pm you’ve cut the 33rd version of this release and you could just push the button to release it – but no, you go through the manual checks just once more. You write just one more test case that’s been nagging at you, you do one more test release on some weirdass architecture, just to make absolutely sure it works.

Why? Well part of that, surely, has to be a fear of ridicule, of being found wanting. Nobody rewards you for doing the same good job day in day out, that’s just life, and personal commitment to quality, no matter how strong, only takes you so far.

I hope that Shuttleworth has a community strong enough to encourage that sort of quality even without the viciousness, but it’s possible that by removing the combativeness of the lists he is castrating some of what makes Debian so remarkably good.

0 Responses to “Shuttleworth on Debian”


Comments are currently closed.