An interesting poll result...

[BlueDaisy]

http://www.pcworld.com/businesscente...x_distros.html

...From a poll by linuxquestions.org. It appears there's a major trouble figuring out userbase and popularity among distros. No one would have thought slackware would have gotten these numbers...

Which makes me wonder: Does anyone has an idea of a statistical source that would best explain the relative popularity and use of a distrobution? And even more, what's that best one measuring the usage of Kubuntu specifically? It would really help, so to see how high/low use of Kubuntu has come up.

[woodsmoke]

The real, and continuing, problem is very simple.

Nobody who "runs" a distro actually has the cajones to just post the number of downloads that they have....that is the ONLY real "number".

woodsmoke

[GreyGeek]

Just like using DistroWatch page-hit ranking to determine the "most popular" Linux distro, the polls are equally misleading. Just like this pole, when you check the numbers you'll see that the poll does not have a valid sampling range to justify the claims made by the pole or Noyes.

It appears the the total number of votes was less than 1,000. That is far to few except in only the best designed and executed polls. Many distro forums have users who "get the troupes" out when they see a poll so that they can influence the outcome in favor of the distro they are using. Unregulated sampling renders such polls totally useless. A better poll would be created by a double blind random sampling of a thousand Linux users by direct contact, not solicitation of voters by some web announcement.

I've remarked about Noyes before.

[woodsmoke]

GG
and that whole thing about "getting the votes out" is SO BOGUS.....and also reminiscent of what goes on in national politics of ANY nation.

woodsmoke

[BlueDaisy]

Yes, I always thought every poll without true random sampling appeared decieving, including this one. But how can we ever find decent facts about distribution usage? I mean even the legitimacy of Canonical's claim of having 20 million users might be inaccurate, in that I heard the number could represent all ubuntu based distributions put together (including linux mint):

http://distrowatch.com/weekly.php?issue=20120206

(the questions/answers corner)

[woodsmoke]

Hi BlueDaisy
yes, I read that somewhere myself, that Canonical included "Ubu" and all the "deriviatives" as "Ubu"...

Linux people are somewhat more "aware" than other computer users....but just a little "more" .... lol, but, I think most of them thought of the statement as a stretch.

woodsmoke

[SecretCode]

The real, and continuing, problem is very simple.

Nobody who "runs" a distro actually has the cajones to just post the number of downloads that they have....that is the ONLY real "number".

woodsmoke

I dispute that. It's much more representative than distrowatch hits, for sure, and more objective than small polls like that at LQ.

But "downloads" != "runs". I've downloaded many distros, and versions of distros, that I don't use after some initial messing around. For desktop usage, counters of user-agent strings of visitors at major websites would be a much better representation of what is actually used ... except that we know user agent strings can be incomplete or spoofed.

eta ... also, n(downloads) != n(installs) - distros that are strong in corporate networks (I assume there are some...) might show 1 download but have 500 installs.

[kubicle]

Also, most distributions are freely available from mirrors (maybe hundreds of them), a "distributor" only has statistics of downloads from servers they control themselves. Better than small polls, but certainly not something to rely on.

[woodsmoke]

I think, that, the participants in this rather intelligent forum should come up with the sine qua non of methodology to figure out the answer to the problem.

woodsmoke

[Detonate]

I think two of the best classes I took in college were Statistics. Ever since I took those classes, I now look at all poll results with skepticism.