Log in

No account? Create an account
ubuntu, wine

Estimating our user base makes me want to contribute more

The Fedora project is claiming between 9.5 and 10.5 million users, as measured by unique IP addresses asking for updates*.  This puts them ahead of the latest published estimate for Ubuntu, of about 8 million users.

The 8 million figure is out of date.  I first remember hearing it about a year ago.  Historically, Ubuntu has been doubling in users every year, so there's a good chance the 8 million figure is far too small.  The 8 million was also a conservative estimate from Canonical using someone mysterious means.  It would be nice if they were open about the process - at the very least we could release a figure using the same methodology as Fedora (unique IPs poking the update server).

I'm skeptical of Fedora's claim to be ahead of Ubuntu, even if the 9.5 million users is accurate.  In 2007, a Desktop Linux Survey estimated Ubuntu had a 30% marketshare while Fedora had about 6%.  A google trends search shows searches for Ubuntu increasing substantially over that time, while Fedora has been decreasing.  It just doesn't seem right that Fedora would quintouple in marketshare while simultaneously experiencing fewer google searches.

Now, people often only run google searches when they have problems; trends data could then be a measure of bad things.  However, it's not a terrible proxy for popularity; Ubuntu is 5 times as trendy as Fedora, which fits nicely with the survey data.  It's unlikely Ubuntu users would be 5 times as likely to google as Fedora users.  Perhaps another desktoplinux.com survey will help settle things.

Usage Statistics are Important:

Every once in a while, I ask myself if my volunteer contributions matter.  I know from popcon results that Wine is installed on about 10% of Ubuntu machines.  I have hundreds of thousands of users.  If I were to improve the Wine package to give the software equivalent of a stick of gum to every user (about 10 cents or so), then in real terms I've just created 80 thousand dollars in value.  That's a conservative estimate: maybe there are twice as many Ubuntu users, popcon doesn't count Wubi users who are probably more likely to use Wine, and so on.

I do this for free.  I've been working on Ubuntu and Wine for about 5 years now, and sometimes when I'm low on money I find myself asking if it's really worth it.  I struggle with a part time job to make rent, and dedicate the equivalent of a full-time job volunteering for Ubuntu; meanwhile, my consultant friends charge 150 an hour to format/reinstall Windows.

I ask myself if it's worth it.  Looking at those user stats, I have no doubt that it is.  Even if I'm only making a stick of gum's worth of improvement.  Sometimes it's a thankless job, but then again charity usually is.

*bad journalism: see first comment

Tags: ,



Not quite

The Fedora Project doesn't make this claim at all. This claim is from the author of the linked article.

Fedora is quite frank about it that these numbers are just unique IPs and don't translate into user numbers, just like being 'trendy', google trends, distrowatch rank, etc. The actual number of Fedora users is unknown to the project, just as Canonical can not know how many people use Ubuntu.

See also https://www.redhat.com/archives/fedora-marketing-list/2008-November/msg00125.html

Re: Not quite

Thanks. The article is really bad journalism in this regard, as it ascribes far too strong language for any of the evidence (or supposed claims) it contains.

On the other hand, unique IPs isn't a particularly awful way of estimating users. Canonical has some difficulty as they likely don't control every mirror, and thus can't get all the unique IPs.

Yes, unique IPs over the past week will overcount users with dynamic IPs, especially dialup users that change IP every day when they reconnect. On the other hand it will probably undercount even more users - those behind NAT (Linux users tend to cluster together), those not requesting updates, and those who use their own internal update server (eg everyone at Google.)

It's completely plausible that Fedora has 10 million users or so based on unique IPs over the past week (though not a substantially longer period). But if that's true, then Ubuntu would have about 5 times as many given last years marketshare, a number over 6 times what Canonical has estimated.


Thanks for all your hard work -despite I don't use Wine since I don't need to run any Windows apps in my comp-!


Thank you

I'm a Wine user and I can say that your efforts are highly appreciated. Thank you for putting your time into Wine and Ubuntu.




I am not really a Wine user, but I recognize that it is a very worthy piece of software. I would really like to thank you for all your work. Some people always give, others never. You make the world a better place, IMHO.


Thank you!

This is just to say Thank you very much for contributing to Wine! I am using Wine to play HL2 and Team Fortress via Steam, and although it is a little bit slower than it was on Win XP, it is really amazing that all this complex software works with this software layer. That is amazing, I really I have to say.

So thank you again and I hope you keep up the good work and find a better payed job, hopefully FOSS related!

Is the 8 million estimate out of date? How can you say that if the methodology which produced that number was never documented or communicated? Please provide me a reference that explains the methodology used to produce the 8 million number for Ubuntu. Shuttleworth is on record stating in a teleconference to the press in the past month that Canonical has no idea what the userbase number is. If that is so, how is that 8 million user estimate derived? Do you has a Ubuntu community member have any idea how Shuttleworth came up with that number? If not how can you say its out dated? Do you know something Mark doesn't?

Reference: The Register Article

Unfortunately the audio of that press teleconference doesn't appear to be publicly archived so they only people who know the full context of the quote are the people who were in on the call.

Every time the 8 million user estimate is used with regard to Ubuntu... is bad journalism. There is no communicated methodology on how that number was created. If you know how it was created... please...tell us. Until the process by which that estimate was created is publicly described, that number can not be independently verified nor can the value of the estimation process be assessed.

I have no idea how it was created, but I can definitively say it's out of date since I first heard it about a year ago, and Ubuntu has almost certainly grown since then.

I'm on your side here: I really do wish Canonical was public about how they're estimating their usage data. 8 million didn't seem that unreasonable a year ago, so I was willing to accept it then. Now, it's likely a lower-bound.
Or it could be completely made up. You have no idea... you can't make a judgement as to the validity of any estimate unless you know how it was derived. Stop putting the cart before the horse and reach up into your own community's leadership and marketing team and demand a communicate methodology before a number is pulled out to the press. Define the process, communicate the interpretation that will be used, then calculate the number. Don't do it the other way around.

And fyi... the desktoplinux survy poll is deeply flawed as a marketshare estimation concept because the polling group was not a random sample. It's really important to understand sampling theory and how critical the concept of a random sample is if you are going to extrapolate the findings and applying them to a larger population.

As for google trends....

google trending ubuntu,windows xp, windows vista is an interesting excersise.


Obviously the way to interpret that is that Ubuntu has equal marketshare with either windows xp or windows vista. Wow! that's awesome. Or then again maybe not. Perhaps its best to just realize that the theory that google trends tells us something significant about market penetration is unproven. We really don't know what Google trends tells us.. the graphs are pretty though. Just because you have a pile of data doesn't mean that all the correlations and comparisons that you can derive from it can be interpreted self-consistently. Fitting your intepretation to just a couple of specific google trends, doesn't tell you the intepretation is a value way to look at how google trends comparisons work. If the result of the ubuntu, windows vista, windows xp trending don't make sense based on your chosen interpretation for other ubuntu questions.. then well.. it doesn't necessarily make sense for the question you care most about either.

Hey but trending is cool. Here's a couple of trend graphs that is really interesting:

So what does that tell me in comparison to the google trend? More people google about Ubuntu.. but more people could get a job associated with Fedora? Hmm...interesting.

So more people google about ubuntu, but massively more people could find a job about linux? Hmmm.

others are just as fun.

What's the take away lesson, want a job? Learn Centos.

I know about valid statistical methods. I see no reason to reject the desktoplinux.com survey out of hand, as we have no reason to believe that it's systemically biased in favor of one distro or another. To do so would be to make a very strong normative claim about the users of one distro over the other.

Desktoplinux.com doesn't pretend to be a "random sampling" -- the population it's studying is deliberately supposed to be users of desktop linux. Not all people, not all computer users, not even all Linux users. Fedora and Ubuntu are primarily desktop distributions, comparing them with the desktoplinux survey is much more useful than comparing them with job trends data.

Also, please don't patronize me by citing how Google trends isn't valid for comparing Ubuntu to Windows. The reason is obvious - Windows users have no reason to specify "windows xp" in their search terms, and frequently have no reason to even specify windows. This isn't the case with Fedora users, and likely isn't the case with Ubuntu users either.

And, yes, Red Hat and Centos are much more popular in the enterprise than Ubuntu. This is well known, and not contested by total usage stats.


Many users either way

It is in Red Hat's business interest as a public company to publicly announce they have a large user base for Fedora and also to say they have more users than x,y,z distros (it was implied in his statement). It should be viewed as a business statement. Fedora is a good distro, and I don't think it really matters if they have 10mill or 7mill, well, it may to Red Hat's executives.

Ubuntu and Fedora has millions of users, both contribute alot, I am very happy with Ubuntu and if Ubuntu has more users, well good, if not, good too, either way as long as there are more Linux users. We need more. :)

But what is the best way to calculate how many users? I haven't registered my Ubuntu with popcon, I've never heard of it until yesterday.


Wine is awesome

You wine people are awesome. Just the other day, a friend of mine ran a Windows antivirus on his Windows partition using Wine from his Ubuntu partition because it was easier than figuring out clamav. And it worked great.

Thanks again.

PS: The antivirus software was Avira, just in case you want to know.

Thank you!!

I am a Wine user so the least I can do is give you a little thank you :)

>> Thank you for all your hard work



it is a thankless job, thank you for carrying the burden. I hope one day to be able to give as much as you have without expecting return. thank you for giving me freedom in my little box because everyday this world takes more and more away.
your a hero.

- helbuns


claim insurance

I have to say that I thought this piece was very profound. I have already bookmark you page. It has shown me a new insight in to my research about current government policy.


Репетитор Химии

IT"S AWESOME!!)) at the last second of this video THERE IS SUCH A FUN)


Dissertation Samples

Yes, I have accomplished my research process after visiting this blog.



ubuntu, wine

April 2009

Powered by LiveJournal.com