Andy Lester wrote:
Don't worry about the game.
But that's all I hear about is the game.
It's all game game game game.
Just because people talk about the rules of the game doesn't mean that
there isn't a benefit to having people playing the game.
Look at anything from Perlmonks to WoW. Metrics and affirmation matter
to people. It's amazingly cool that we can get people to change what
they do because they want their metric to go from X to X+n.
The talk about the game is trying to reconcile the fact that "quality"
is not something that is easy to quantitatively define. But that's OK.
As some famous (likely dead) statistician said, "All models are wrong;
some models are useful".
There's not even a real downside to people "cheating" at this game, as I
don't think quality can go down from people cheating. All we really
need to do is ensure that new rules don't (a) make people disregard the
game or (b) make people lower quality to improve scores.
And so we talk about the game and the rules and whether we want to add
more rules or change existing ones. As long as some people -- for
nothing -- might be incented to fix problems in their distributions,
then it seems like a net positive to me.
Has anyone done any trend analysis over time on the quality
distribution? Normalizing for new quality point that have been added,
has quality gone up measurably since CPANTS began? Or put differently,
if we tracked the percent of CPAN modules that got the quality point for
each of the metrics, how would that graph trend over time? More "use
strict"? More "META.yml"? More "README"?
Regards,
David Golden
