On Mon, Jan 28, 2013 at 09:13:33AM -0700, Robyn Bergeron wrote:
On Mon, Jan 28, 2013 at 8:26 AM, Jiri Eischmann
> frankly I'm not sure which mailing list I should write this to, but I
> guess the marking one is the best fit.
> I was really surprised (in the negative way) by the Fedora stats .
> Repository connections in the first week dropped by 52 percent compared
> to F17 and F17 was already low compared to releases before. F18 has
> pretty bad reception, but such a huge drop is hard to believe.
> The number of direct downloads is down by 7 percent.
> Are these numbers still comparable or the way they're calculated has
> changed? If not, it's a pretty bad trend.
So a few things -
This script still runs in Paul's fedorapeople space pending the
arrival of the glorious "mustard" application (note: name not actually
determined, but I'm pretty sold on that name) which will replace
/Statistics. It occasionally misbehaves, and Paul sent a note saying
that there was a bit of a hiccup, but we should be back to normal
number reporting next week, but this is the number he had for last
week (I'm unclear about if the current number in there is correct,
totally off, or what.)
The statistics mail normally gets mailed out on Monday nights late, so
perhaps tomorrow's data will look more normal (or more alarming).
WRT direct downloads: I've noticed that even-numbered releases tend to
be on the downward trend against the previous release in the first few
weeks, whereas odd-numbered releases usually are more than 100% of
Paul: can you shed any additional light?
There was a minor hiccup, in that the direct download number was not
reported correctly. I believe I've fixed that now, and I was able to
re-run that statistic so the number on the wiki should be as expected
now. It should be easy to tell from tonight's report whether the fix
The number for yum repository update connections comes directly from
statistics compiled by a script provided by the Websites team (a cron
job called zFedoraUsage, IIRC). I don't know whether they've switched
their methodology for this, but it certainly would make sense to ask.
I'm cc'ing Kevin Fenzi for this.
Paul W. Frields http://paul.frields.org/
gpg fingerprint: 3DA6 A0AC 6D58 FEC4 0233 5906 ACDB C937 BD11 3717
- - - - http://pfrields.fedorapeople.org/
The open source story continues to grow: http://opensource.com