On Tue, Nov 02, 2004 at 04:37:25PM -0500, Joey Hess wrote:
> Eduard Bloch wrote:
> > to visualize what we all already know, I created a simple ranking page
> > that counts the bug reports of a certain person together and gives it
> > appropriate weight (based on the priority of the package, the time the
> > bug has been kept open and its severity). The results can be found here:
> >
> > http://people.debian.org/~blade/pranger/ranking.html
>
> I seems to me that by these metrics the "best" way to contribute to
> Debian is to find something extremely obscure that has 1 user (you, or
> perhaps your cat), and a dead upstream, and package it, possibly with a
> description like "this package is experimental, not very useful, takes
> six months to learn, and you should use bar instead". You then get 0
> bugs and rise right to the bottom of the list.
Yes, this metric basically sucks. I created a better one a while
ago[0] (I know that script is around here somewhere...), based simply
on the length of time which RC bugs have been left unattended. While
it had a number of flaws in the simple form, they would be easily
correctable. It went something like this:
score := sum of ages in days, for all bugs over 30 days
And then delete all scores below 50 or so. So one bug unfixed for two
months, or two bugs unfixed for a month, would qualify. I fiddled with
the numbers a bit until I got a convincing result, based on inspection
of the top and bottom ends of the list.
Unattended non-RC bugs may be interesting, but they just don't
correlate. I tried; there's no appreciable relation. So you have to
ignore them. There *is* a significant relation with old RC bugs, once
you strip out the noise generated by NMUs and WNPP stuff. Package
priorities have no relevance.
I had been intending to do something more regular... but damnit, TIME.
[0] And it worked as intended, too. A significant number of the people
near the top of the list I posted sorted their crap out. The fact
that it was simple and unarguable probably had a lot to do with
that: if you had a high score under that metric, you clearly
sucked, and the only reasonable responses were to do the damn work
or find somebody else to do it. Of course, it didn't work for very
*long*, but...
--
.''`. ** Debian GNU/Linux ** | Andrew Suffield
: :' : http://www.debian.org/ |
`. `' |
`- -><- |
Attachment:
signature.asc
Description: Digital signature