I read a little bit of the most recent humongous thread on the devel list. For those who like to ignore threads with 70+ replies in them, it’s about changing the Fedora updates system to not allow anyone to go straight to stable. (The policy isn’t actually written yet, so we don’t know if packages will require testing or if we can wait until the old_testing period is up or what. I also just summarized 70 emails worth of data into two sentences, so obviously I’m missing something—if you want to bite, read the thread.)

I have never had karma posted to a testing update—positive or negative—unless I, well, basically campaigned for people to test and vote, and I only did this for Gwibber when I maintained it. I spent more time trying to get three people to test this update than I did rebuilding and testing the update myself. I really only did this because I thought it was for the betterment of Fedora, and it really is, but it was one of the reasons I gave up maintenance of Gwibber: I didn’t have enough time to test it properly myself. (Then again, I had stopped using it for personal use as well.)

Gwibber is one of the few packages that I pushed to testing, and I only started doing so after someone started asking maintainers why they were pushing straight to stable. I didn’t have an excuse. :)

The bigger picture is that people don’t have time to give positive karma on packages that aren’t actively tested (like kernel). I’m going to source some conversation I looked at in #fedora-devel (and hopefully the owners of the words won’t care):

< hno> I usually keep my updates in testing until getting that email that it's
       about time to push it, which I for a long time took as a policy. But sad
       truth is that the amount of feedback given from users is usually nil..
< hno> and without users testing it does not really matter what policy there are
       imho.
< nirik> well, no karma does not mean no one tested it.
< nirik> it means that it didn't break or annoy them enough to go -1. Or matter
         to them enough to +1
< hno> true, but it also means the packager have no idea if it's been tested.
< kylem> the kernel has a huge problem here. nobody notices it's running unless
         it doesn't work.
< kylem> that's not as big a problem as the people who don't understand what
         regression means.
< kylem> but just vote -1 to every kernel because i didn't bother to fix their junk.
< kylem> which actively harms everyone else.
< kylem> makes me a sad panda.

I don’t know how to refactor the updates system so that we have more effective package updates and more effective testing. I’m no updates engineer. But I do think that requiring all packages to go through a two-week or indefinite testing period really pushes (hah) us farther and farther away from our why upstream policy.