faq
flatforty
contribute
subscribe
configure
search
rdf
main
parent
thread
|
Re: Desktop Linux Survey
by Anon on Thursday 23/Aug/2007, @00:35
|
I was actually logging the total number of votes and percentages of GNOME vs KDE every 5 minutes over a long period, and the logs seem kind of ... well ... odd, to me.
Does anyone know if there was any checking for non-unique IPs in this survey?
The general trend was for KDE's share to increase and GNOME's share to decrease but then the number of votes/ 5 minutes would spike suddenly (by an order of magnitude) and these votes seemed to be almost *all* for GNOME. These spikes happen twice (actually bringing the survey server down at one point) and also oddly, the shares of OpenSuSE, Firefox and Evolution all increased hugely during them (I don't have these in my logs, though), which is doubly odd as KDE is dramatically preferred by OpenSUSE users. After the spikes, the trend of increasing KDE and decreasing GNOME shares resumes again, but the flow of votes is at a trickle by this stage, so there's no hope of catching up.
If anyone knows of a site that allows anonymous uploading/ sharing of text files, I'd be happy to share the logs. This probably sounds like tinfoil, but the huge spikes of OpenSUSE+GNOME+Firefox+Evolution votes are fairly dubious.
It would be nice if, in future polls, Captchas were used to make the results more trustworthy. |
|
|
The Fine Print: The following comments
are owned by whomever posted them.
( Reply )
|
Re: Desktop Linux Survey
by Alistair John Strachan on Thursday 23/Aug/2007, @00:52
|
I think the results are fairly logical given the lead Ubuntu has on other distributions at the moment. Kubuntu is simply a minor distribution by comparison.
Personally, I think KDE is doing very well to hold the minority it has. Clearly, if you discount Ubuntu (for which most new users are not presented with a facility to install KDE, or even know what it is), KDE is doing very well.
|
[
Reply To This | View ]
|
Re: Desktop Linux Survey
by Boudewijn Rempt on Thursday 23/Aug/2007, @01:58
|
"Does anyone know if there was any checking for non-unique IPs in this survey?"
Nope.
|
[
Reply To This | View ]
|
Re: Desktop Linux Survey
by Sebastian Kügler on Thursday 23/Aug/2007, @03:20
|
Nope as in "no one knows" or "there isn't"? :-)
The results look a little suspect to experiences I've seen in other fields. If 25% of the total number of votes have indeed been cast in only one night, it looks like someone wanted to screw up the results. On the other hand, it's easy to interpolate in that case (but probably no one will). It doesn't do the value of the survey any good (but then online surveys are always subject to this kind of fanboy manipulation, which is why they should be taken with a grain of salt in any case).
"There are lies, damn lies and statistics" as some smart individual once noted. A nice, but unrelated site that deals with what's hidden behind numbers is freakonomics.com, by the way.
|
[
Reply To This | View ]
|
Re: Desktop Linux Survey
by Boudewijn Rempt on Thursday 23/Aug/2007, @05:13
|
There isn't -- I was a little surprised by one of the spikes myself and tried to enter a second vote, just to check whether they checked. Since I succeeded, I assume that at least at one point in time there were no checks at all.
|
[
Reply To This | View ]
|
Re: Desktop Linux Survey
by jospoortvliet on Thursday 23/Aug/2007, @04:40
|
as far as I could see, they did at least SOME check, as I couldn't vote for both myself and my ex (I use konqi, she uses firefox, so I thought I should mention that in the survey) from home...
|
[
Reply To This | View ]
|
|
Re: Desktop Linux Survey
by j.v. on Thursday 23/Aug/2007, @02:35
|
Between Thursday and Friday the total number of voters increased
by about 10000 voters, all or most of which were OpenSUSE+GNOME.
I wonder why desktoplinux.com doesn't bother to also publish a graph
of their traffic sources...
|
[
Reply To This | View ]
|
Re: Desktop Linux Survey
by Inge Wallin on Thursday 23/Aug/2007, @03:49
|
I think it would be extremely interesting to see these statistics.
We must realize that this survey is going to be touted as proof that
Gnome has now "overtaken" KDE by both ignorant people and people with
an agenda. It would be very valuable if we can set things straight
and perhaps also get a comment from DesktopLinux.com.
Right now, there is an "analysis" stating exactly what I say here
above, and that irks me since I think that it's false. Naturally,
it's also detrimental to us if people think that we're being
abandonded.
If you want to publish your figures, then I am willing to do a writeup
complete with nice charts that make the issue as clear as I can.
|
[
Reply To This | View ]
|
Re: Desktop Linux Survey
by EP on Thursday 23/Aug/2007, @05:46
|
"If anyone knows of a site that allows anonymous uploading/ sharing of text files, I'd be happy to share the logs. This probably sounds like tinfoil, but the huge spikes of OpenSUSE+GNOME+Firefox+Evolution votes are fairly dubious."
Yeah. Please share your logs. Wouldn't something like http://pastebin.com or Rapidshare not suffice?
Of course even if outright manipulation did not take place, the survey results would be almost worthless nevertheless. As a minimum measure to estimate the real figures one would have to find out how inclined the specific user groups are to take part in this survey.
Even if reliable figures can be obtained, there is still the question how relevant the results are. For example, it was only asked "which web browsers do you frequently use", it was not asked how often, and more important, how long the browser is really used. It is possible for instance that non-frequent users just go with the defaults, whereas frequent users deliberately choose another browser.
Also, usage preferences may not be mistaken as a measure of the quality of the programs. This I think is the most important point for KDE developers and users. Consider this: The question for web browser preferences is actually skewed *against* high-quality desktop environments, since in a well-integrated desktop environment like KDE, you do not need the browser so often, since you can just use http-urls in the applications themselfs (KPDF for example).
|
[
Reply To This | View ]
|
Re: Desktop Linux Survey
by bedalius on Thursday 23/Aug/2007, @05:51
|
You can upload your file at www.rapidshare.com
<I>This probably sounds like tinfoil, but the huge spikes of OpenSUSE+GNOME+Firefox+Evolution votes are fairly dubious.</I>
Why? Sounds like Nat Friedman pushed his favorites...Just kidding ;-)
|
[
Reply To This | View ]
|
Re: Desktop Linux Survey
by sjvn on Thursday 23/Aug/2007, @06:26
|
> Does anyone know if there was any checking for non-unique IPs in this survey?
Yes, we checked and blocked non-unique IPs from non-corporate IP ranges and we tracked corporate use IP ranges. This prevented Joe Home Linux Fan from voting over and over again. We also made sure that JoesLinux.com couldn't stuff the ballot box for JoesLinux.
The data made it pretty clear that the GNOME bump came from Ubuntu users.
FWIW, as a dyed in the wool KDE user, I think they don't know what they're missing.
Steven
|
[
Reply To This | View ]
|
Re: Desktop Linux Survey
by Konqi on Thursday 23/Aug/2007, @09:44
|
> The data made it pretty clear that the GNOME bump came from Ubuntu users.
Erm, how is this supposed to work mathematically? During the 10000 vote "peaks" in question openSUSE votes had to increase up to 40% for the whole duration of the peaks to get 20% for the total vote. During that time the amount of Ubuntu users must have _decreased_ actually. So I can hardly see how you can attribute the change of the poll to the Ubuntu users.
> FWIW, as a dyed in the wool KDE user,
> I think they don't know what they're missing.
We know that you like KDE. You know it's neither meant as a rant, nor a personal attack :-)
|
[
Reply To This | View ]
|
|
Re: Desktop Linux Survey
by Phase II on Thursday 23/Aug/2007, @15:14
|
Thanks for bringing this up. You're not the only one who has caught mysterious shifts in the vote results. Some days ago, triggered by another comment and having the results from the day before in mind, I checked about two days and three days after vote start and already saw a noticably shift in a certain vote constellation.
see here http://dot.kde.org/1187268289/1187269736/1187272356/1187273173/1187274305/1187288798/1187331858/1187343466/1187344416/1187353058/1187362815/
After about 2 days the results were about settled (a number people from most continents/timezones having voted so that results weren't easy to skew anymore) so it could be expected that only total vote counts would go up after that.
But still, one vote constellation (which was: openSUSE, Gnome, Firefox, Evolution and "None -- I don't run Windows apps on Linux") would rise noticably after these about 18.000 votes. After the final 38.500 votes the results shifted like this:
Vote option | percentage after about 18.000 votes (after about two days after start; sorry, only rough numbers) | percentage after about 27.000 votes (half a day later) | final percentage after 38.500 votes
openSUSE ~9,5 ->16,5 -> 21
Gnome ~37,5 -> 42,5 -> 45
Firefox ~54/55 -> 58,5 -> 60
Evolution 14 -> 21,5 -> 32
None -- I don't run Windows apps on Linux ~22 -> 30 -> 39
To summarize, it was as if almost nobody voted for any other option anymore; massive shifts, even percentages which more than doubled!
Even for persons who are not familiar with distributions of votes and normal shifts which may occur, it's not hard to see that this is very, very, very unlikely. Or to summarize, the poll has been manipulated.
That no suspicious votes have been seen by desktoplinux.com doesn't mean there haven't been any.
(Perhaps someone can calculate the total number of votes which exclusively went to that combination. As the other vote options didn't change drastically, mostly only went down, I didn't track these.)
Sadly this poll is clearly manipulated and cannot be taken as a reliable source for the percentage a distro or application has on Linux. As I already wrote in the first comment, I also wouldn't trust these kinds of votes if it hadn't deliberately been manipulated.
P.S.: I don't really care for the manipulator (aka the one with the bot net?), but, well, I don't think he did any good for the options he pushed. His work is obvious, so is he really helping this certain distro, desktop, application or attitude he voted for? Perhaps publicity, but as it is so plain and simple obvious, it's quite certain that people will point at it, which reverses to bad publicity. The most certain thing one can say is: no reliable numbers this year.
|
[
Reply To This | View ]
|
Re: Desktop Linux Survey
by Div on Thursday 23/Aug/2007, @16:27
|
Honestly, your numbers are poor and mean nothing.
|
[
Reply To This | View ]
|
Re: Desktop Linux Survey
by André on Friday 24/Aug/2007, @01:32
|
Since you seem to have better insight into this than the rest of us: please share with us why you think these numbers are poor and mean nothing. Better yet, please explain the observed shifts yourself.
|
[
Reply To This | View ]
|
Re: Desktop Linux Survey
by jospoortvliet on Monday 27/Aug/2007, @02:21
|
I've asked some ppl here (I work in a research institute) about how to analyze the results with statistical software, and see if it's significantly wrong. So I told them about the results (25.000 votes in 3 days, so on average a little under 350/hour. Then Friday night, from 3 to 4 o'clock - 3000 votes, 2700 Gnome+Suse+F etc, 300 random around KDE. How do I analyze this, see if it's likely to be random or not.
They said you don't need to analyze that, a child can see such numbers aren't normal.
|
[
Reply To This | View ]
|
Re: Desktop Linux Survey
by theriddle on Tuesday 28/Aug/2007, @15:07
|
Agreed. Any brainless 2-year-old would know these #s are unreal.
Unfortunately, no-one else would bother to check out and will gloat about how GNOME has overtaken KDE. The fanboy who screwed the results will get what he wanted.
|
[
Reply To This | View ]
|
Re: Desktop Linux Survey
by Carewolf on Friday 24/Aug/2007, @02:03
|
So lets analyze the numbers.
openSuse:
18000 * 9,5% = 1710
27000 * 16.5% = 4455
38500 * 21% = 8085
Differences:
18000 early votes: 9.5% openSUSE
9000 middle votes: 30.5% openSUSE
9500 late votes: 38.2% openSUSE
GNOME:
18000 * 37.5 = 6750
27000 * 42.5 = 11475
38500 * 45 = 17325
Differences:
18000 early votes: 37.5% GNOME
9000 middle votes: 52.5% GNOME
9500 late votes: 61.6% GNOME
Strange, but no smoking gun. You need time stamps to prove the "10000" strange votes.
|
[
Reply To This | View ]
|
Re: Desktop Linux Survey
by scratching my head on Friday 24/Aug/2007, @09:44
|
The News page at opensuse.org ( http://news.opensuse.org/ ) has this to say about the survey:
### . What’s striking is that popularity of openSUSE almost
### . doubled form (sic!) roughly 10% in 2006 to close to
### . 20% in 2007 survey.
Oh yeah, really striking...
I'm an openSUSE user myself, and I like the distro quite a lot, as well as their fine developers (I don't like their upper/top management types though).
But with these DesktopLinux figures, something smells fishy....
And also compare them to the figures that Novell's own recently published user survey showed: KDE there was by far the most popular desktop environment still, dispite the lots of "anti-KDE" bruhuahua that seemed to originate from their management level for a certain period....
|
[
Reply To This | View ]
|
|
The Fine Print: The previous
comments are owned by whomever posted them.
( Reply )
|
|
![[KDE Dot News]](http://www.kde.org/dot/Images/kdedotnews.gif)
