Of hall monitors and slippery slopes

Posted May 12, 2010 18:51 UTC (Wed) by lmb (subscriber, #39048)
Parent article: Of hall monitors and slippery slopes

Testing is paramount to quality. If it wasn't tested, it is broken. There is way too little testing in the software world. Test-driven development would be such a good idea.

What may be needed is a meta-vote - essentially, a review by high-karma users not of the change or the package, but of the package's automated regression test-suite. A package with an excellent regression test suite could get away with shipping the change quickly.

And, possibly, maintainer karma - and repeated demonstration of excellence on updates - could simplify and speed up the release process.

to post comments

Of hall monitors and slippery slopes

Posted May 12, 2010 20:22 UTC (Wed) by mrshiny (guest, #4266) [Link] (2 responses)

Karma could be granted (or removed) afterwards as well, based on the number of bugs reported in a particular update (in the absence of a regression test, the users are the regression testers). In this way over-aggressive package maintainers or problematic packages could be identified.

Of hall monitors and slippery slopes

Posted May 12, 2010 23:39 UTC (Wed) by jcm (subscriber, #18262) [Link] (1 responses)

The "number of bugs" is always a bad metric. You could have ten bugs about typos in man pages vs. one whopper - how do you measure that?

Of hall monitors and slippery slopes

Posted May 13, 2010 3:38 UTC (Thu) by mrshiny (guest, #4266) [Link]

Well, presumably the bugs can be weighted by their severity (assigned by a person) or by the number of times this bug was reported (by the automated reporting tool). Maybe "number of bugs" isn't accurate but rather "number of people who reported this" or "number of crashes caused by this bug".

It isn't perfect but it's a start.