Re: The number of popcon.debian.org-submissions is falling
[Jesús M. Navarro]
> I for one don't allow for popcon on production machines under my
Sad to hear that your machines are not able to influence the parts of
Debian which uses the popcon.debian.org numbers to improve Debian
(like the CD/DVD builds :).
> I'm not going into details about why this is the case now, but
> that's the fact (on a side note I'd want to know if there's any easy
> way to mimic/modify Debian's server/client popcon enviro so it can
> be used to produce my own internal stats).
I did this for Debian Edu (<URL: http://popcon.skolelinux.org/ >), and
the scripts to set up your own server is in the popularity-contest
package, in /usr/share/doc/popularity-contest/examples/.
The clients can be configured to submit reports to several URLs, and
you can then get it to report to your own collector or the debian
collector and your own collector (like we do with Debian Edu).
> If that's the case with other sysadmins that probably would mean
> that main source of popcon machines are home systems which in turn
> would make understandable for popcon numbers to decrease by the
> "EOL" of a Stable release:
Given the amount of server packages listed as used in
popcon.debian.org, I doubt it is a general trend. :)
> 1) Current reduction from April onwards is statisticallly
> non-significant with a main trend steadly growing since Aug, 2007.
Could be. Hard to predict the future. :)
I suspect it is a significant reduction, but I do not know the cause.
Perhaps Debian have a reduced user base, people moving to other
distributions. Perhaps the popcon.debian.org server is overloaded and
unable to handle more requests (unlikely, only ~14k submissions via
http every day. Perhaps some strange rumor have spread that popcon do
not work on laptops, and more and more laptops are being deployed in
the Debian community. Or perhaps something completely different. I
do not know, but hope that bringing it up here might get more eyes to
look at the issue and perhaps find a way to get more popcon
> 2) It is i386 the one that basically is stucked which, again, it's
> no wonder since the vast majority of general-purpouse CPUs are amd64
> now and, contrary to 'etch', 'lenny' is quite an usable 64 bit
Well, the overall count is sinking, so that would mean that the "lost"
i386 machines are not replaced with amd64 machines. :)
> 3) Being both 'Y' axis related to number of installs on (mainly, I
> expect) general purpouse systems, I have to wonder why there are
> different scales for i386 and "overall". It takes some time to
> understand there are *not* as many i386 systems as overall (with or
> without i386?) in the popcon and anyway it's impossible to know
> which 'Y' axis belong to what arch.
Sure, the graphs could be made easier to read, but that will have to
be the topic of a different thread. :)