Why is OpenOffice.org so buggy?
February 13th, 2006Andrew Brown’s piece in The Guardian If this suite’s a success, why is it so buggy? raised a minor storm of protest in open-source land at the end of last year. He argued that the OpenOffice.org demonstrated that one of the alleged benefits of open-source is in fact a myth: if a user spots a bug, they can download the source and fix it. Brown argued that the number of hackers as a percentage of OOo users is tiny; the number of hackers skilled enough to attack a huge program like OOo is minute. So, bugs don’t get fixed.
I’ve argued before that the work done by the OOo community is immense, particularly in translations – what’s the use of having the best office productivity suite in the world if you can’t understand it? A posting yesterday by Andre Schnabel gave some interesting statistics about how the community is also helping in the QA process. Starting with the first stage in the QA process – bug reporting:
So, during 1.x codeline most of the issues (87%) have been reported by Sun. For 2.0 codeline majority of the issues (75%) has been reported by the community.
The next stage in the process is triage – confirming/classifying bug reports:
The work of the community with “filtering” and confirming issues is welcomed and of increasing quality.
Coding the solutions is more challenging:
At the moment, the rate of CWS approvals by the community is low, but we hopefully can improve that as well.
So, there is some truth in Andrew’s claims – if you’re an apprentice hacker, we wouldn’t recommend starting with a mature and complex codebase like OOo (although Google’s Summer of Code showed that newcomers could make significant contributions with good mentoring).
But there are lots of other ways in which non-technical people can ensure OOo does remain the leading international open-source office productivity suite. This isn’t just a theory from open-source proponents – it is actually happening in practice. And we can monitor other open-source initiatives such as Ubuntu’s testing process to find out what works and what doesn’t.
