Nice, but it needs more numbers

Story: A Challenge To Canonical, Red Hat, Novell, Linspire, and All Linux Distributors: Put Up or Shut UpTotal Replies: 24
Author Content
Sander_Marechal

Nov 28, 2006
2:06 PM EDT
A nice read, but it needs more numbers. Counting desktop Linux usage is *very* hard I agree, but some reasonably founded guesstimates can be made. Certainly for the bigger distro's. That would be way better than simply stating "it's at least as much as Macs*".

Just tallying up conservative guesstimates for the big distro's you named that should take part in a consortium like this should yield some impressive numbers. For example, *buntu machines are configured to hit Ubuntu's ntp server on startup. Counting hits or IP addresses there could give you some insight in the ammount of Ubuntu installs out there. Similar tricks can be applied to other distro's.

Inaccuracy isn't that much of a problem. An order of magnitude or a ballpark figure is all you're after here, not exact numbers. Also, because you already know the server side you can concentrate on ballpark figures for desktops. This means you can make some assumptions to make counting easier. E.g. quite a few desktop distro's have an auto-update feature, meaning they'll hit the package repository indexes once a day or once a reboot.

* Related: How is it a hardware vendor can technically support Mac but not Linux? It's nearly the same codebase!
dcparris

Nov 28, 2006
2:36 PM EDT
> How is it a hardware vendor can technically support Mac but not Linux? It's nearly the same codebase!

That's my thinking.
rijelkentaurus

Nov 28, 2006
3:05 PM EDT
I've left threads on the forums of Mepis and PCLOS, and I sent an email to the CentOS devs. Can't hurt to try something.

Here's an interesting thought: What percentage of PCs that were sold with Windows now have Linux on them?

What percentage of Macs that were sold with OS 9 or X now have Linux on them?

There's so much out there that is running Linux that we take at least a few numbers away from those two camps, probably more from the first bunch.
tuxchick

Nov 28, 2006
4:45 PM EDT
Sander, those are great points.
Abe

Nov 28, 2006
6:23 PM EDT
Best estimate can be obtained from Internet sites. If every site conduct a tally of their visitors (unique IP, etc...) and publish them on daily basis, we could have very accurate percentage account. Similar to what http://www.webreference.com/ does for web browsers' statistics.
Sander_Marechal

Nov 29, 2006
2:20 AM EDT
I've been doing some more thinking. Is there any data available on how many desktop PC's there are running in the world in total? We could use that figure to derive how accurate the per-distro stats need to be.

For example: say that there are 1 billion desktop PC's running world-wide. We're only interested in giving a percetage figure (e.g. Linux desktop marketshare is 12% as opposed to 12.345%) so we need to calculate total Linux desktop usage in steps of 10 million.

Let's apply the 90-10 rule. 10 percent of the distro's make up 90 percent of the installed base. This matches up pretty well with the distrowatch stats (10% distro's create about 90% of the hits). Distrowatch counts about 500 distro's at the moment so our calculations could include up to 50 distro's or thereabout. We're already guessing that the end result will be around 5% (50 million desktops). 50 distro's spread over 50 million pc's. I think that counting in steps of 500.000 will do fine.

So, what we want then is a per-distro average installed base rounded to the nearest half-a-million for the top 10% distro's. Anything more detailed is just fiddling in the percentage-points and not interesting for our tally. It sounds doable to me. Add it all up, add 11% (90-10 rule reversed) and we got a ballpark figure for the total ammount of Linux desktops out there.

So, anyone know how much desktop PC's were running in 2006 worldwide?
hkwint

Nov 29, 2006
3:16 AM EDT
A while ago, I made a (conservative) estimate of the number of Linux PC's worldwide:

http://counter.li.org/estimates.php

My guess was 29 million users worldwide, probably even more by now.

However, it is VERY difficult to specify who's a user and who not: If you use Windows, but your internet goes through a Linux gateway, does that count? Do PDA's count? What about mobile phones? What about routers? When is somebody a Linux user by the way? If I have a dual boot, then what? If I use a LiveCD or RescueCD sometimes, then what? If I try Linux three months, but switch back to Windows, then what? If I build a small home network with a gateway (so all have the same IP), and run Linux on all my PC's, then what? Do 'servers' count? If I run Cedega or Linux in VMware, does that count? If I run Win4Lin in Linux all day, how would that count?

Indeed, we should estimate the "number of zero's" in the values. 29 million is a bit low probably, but not unreasonable.

Then, don't forget two important projects: -The OLPC project (Fedora) -The Intel / Linux / China project, "Leaked memo from Intel shows Major Linux Effort" http://lxer.com/module/newswire/view/46808/index.html (Don't know what distro)

Also, a lot of Asian people don't visit our 'western' sites probably, so they are not being counted in most statistics we use. For example, looking at some Alexa (or whatever) statistics, the percentage of Asianux will probably not be representative.
rijelkentaurus

Nov 29, 2006
3:23 AM EDT
>My guess was 29 million users worldwide, probably even more by now.

In comparison, Apple claims approx. 15 million...it's somewhere on their website.
dinotrac

Nov 29, 2006
3:29 AM EDT
> Related: How is it a hardware vendor can technically support Mac but not Linux? It's nearly the same codebase!

There may be a matter of perception, or Mac users may buy more stuff. Apple may provide a certain degree of support. Mac OS's are certainly fewer in number than Linux variants. Etc.

The real issue may not be the code base so much as quality assurance. You need to test what you support. That's a much smaller universe in Apple land than Linux.
Sander_Marechal

Nov 29, 2006
4:17 AM EDT
@hkwint: I'm just trying to count plain-jane desktop linux machines here. All the options you list hardly affect the statistics and will be picked up by the +11% at the end anyway.

Afterall, the goals is to show some numbers to the run-off-the-mill hardware manucaturer that now only gives out Windows drivers. Most of what we are clamoring for is stuff make for plane-jane desktops (video cards, NICs, sound, usb devices, music players, camera's), not stuff for high end servers or appliances :-)
hkwint

Nov 29, 2006
5:51 AM EDT
Quoting:I'm just trying to count plain-jane desktop linux machines


Ah, so was I when I said 29 million. Nonetheless, more people use Linux than that 29 million, which was the point I tried to make. If there are 15 million Mac users (didn't know that), then it's indeed strange and not justified 'they' have more drivers. Moreover, the number of Linux users is growing faster than the number of Mac users, as far as I can tell. Probably, those Mac-people (and Apple) have more money or what?
dinotrac

Nov 29, 2006
6:32 AM EDT
>In comparison, Apple claims approx. 15 million...it's somewhere on their website.

That number seems too low, but maybe...

I've seen things indicating that the 16% of computer users use Macs -- ie, installed base instead of market share.

If there are only 100 million computer users, that would be about 15 million, but I suspect 100 million is a tad low.
Sander_Marechal

Nov 29, 2006
6:45 AM EDT
I suspect that too. I was freewheeling on a billion (PC's, not users) above but I have no way to verify that. Anyone know of a source as to how many desktop PC's there are out there approximately?

@hkwint: yeah, your 29M sounds very conservative. Looking at your current stats it looks like the number of Linux users actually decreased (141K noew as opposed to 180K in 2005 registered users) but your page doesn't say how many accounts were deleted since your 2005 estimates.
jdixon

Nov 29, 2006
7:39 AM EDT
> If there are only 100 million computer users, that would be about 15 million, but I suspect 100 million is a tad low.

Are we measuring computers or computer users? They won't be the same.

There are two computer users at our house, but we have 3 active computers in use, another one that gets intermittent use, and probably a dozen in total, most of which have been retired but not thrown out. Those include things like a Commodore 128, a Tandy Model 4, several Tandy Color Computers, a PDP-11, and an Aquarius. It doesn't include the VMPlayer based copy of Ubuntu on my wife's Windows 2000 machine or her Nokia 770.

I think (from memory of quoted figures) that computer usage in the US has topped 50%, by most reliable measurements. With a population of ~300 million, an average of approximately 4 persons per family, and guessing about 1.5 computers per computer using family, that gives ~112.5 million computers in the US alone, and probably 150+ million computer users. Europe would probably be similar percentage wise.
dcparris

Nov 29, 2006
7:57 AM EDT
In that case, Apple's stats would be kind of accurate, at least for the U.S. I would put desktop Linux users at roughly close to Apple's stats, and possibly higher. You can use distro listserv totals to get a base number, but that would ignore those who use Linux, but don't subscribe to the list. You also have to account for the number of people who subscribe to multiple distro listservs. But we could at least get a basic idea of how many people actively use Linux (more accurately, participate in the Linux community).

Firefox and OpenOffice.org also use their download stats, with the acknowledgement that it is not entirely accurate. Still, one gets a rough idea of the number of users.
dinotrac

Nov 29, 2006
8:27 AM EDT
Rev -

I would expect Apple to have more users because they have been selling viable desktop machines for longer than there have been viable Linux desktops.
Abe

Nov 29, 2006
8:40 AM EDT
Quoting:they have been selling viable desktop machines for longer than there have been viable Linux desktops.


Dino: Do you really believe that?

Apple have been selling better desktops a lot longer than MS, how come they are not exceeding MS?
dinotrac

Nov 29, 2006
12:25 PM EDT
>Apple have been selling better desktops a lot longer than MS, how come they are not exceeding MS?

Come on now, Abe. To not believe that, I'd have to be an imbecile.

The first MacIntosh came out in 1984. Linux 0.01 was released in 1991. KDE 1.0 came out in 1998. GNOME 1.0 -- after that, I don't remember when.

As to Microsoft v. Mac, it's pretty easy to figure out: The first IBM PC came out in 1980, 4 years before the first Mac. By the time the Mac came out, millions of machines were in use running DOS and Lotus 1-2-3, WordPerfect, etc.

The very first Macs languished on shelves, unwanted and unloved, probably because they weren't cheap and couldn't do much of anything that a PC couldn't do as well or better. Or, at least, that was true until small laser printers, Postscript, and desktop publishing.

Windows had a leg up in corporate environments, Intel PCs were cheaper than Macs, and, as its market share grew into a monopoly, Microsoft became increasingly ruthless. No mystery there.

hkwint

Nov 29, 2006
1:32 PM EDT
Quoting:but your page doesn't say how many accounts were deleted since your 2005 estimates.
Well, it's not my page, but anyway. The decreasing numbers is because their system changed from 'valid infinitely' to 'must login every year'.

Quoting:I've seen things indicating that the 16% of computer users use Macs
That must be European or American things. Western centric stats, and Apple propaganda which is not representative for the whole world. Do you really believe 16% of the Chinese farmers, or poor people in Brazil or Africa (and other poor people not living in the west) - if they have a PC at all - have money to buy a Mac? Some will own one anyway, but 16% is not realistic at all.

A while ago I read the billionth PC was sold. If half of them are broken, there are still 500 million of them, though I have no idea how many of the sold ones still remain. A lot of them are probably not able to run Windows XP.
number6x

Nov 29, 2006
1:35 PM EDT
"In comparison, Apple claims approx. 15 million...it's somewhere on their website"

I think that would be the numbers of macs sold per year. The number of users would be greater, sinc most macs(except the Performa) are used for more than a year.

:)

hughesjr

Nov 29, 2006
1:50 PM EDT
I can tell you that CentOS has had more that 1 million Unique IP addresses do yum and up2date updates against our "*.centos.org" servers in the last 6 months.

Those servers only do about 33% of the updates (140 external public mirrors do the rest).

That means about 3 million unique IPs for CentOS.

Now this number is high because many people have dynamic IPs, but it is also low because many companies and ISPs run their own update servers and never hit any CentOS servers ... so I would call those factors a washout.

We do not ship proprietary drivers or patent encumbered software.
Sander_Marechal

Nov 29, 2006
2:25 PM EDT
Thanks hughesjr. Now, if the distrowatch stats match realworld installations in any way (highly unlikely) that would put the total ammount of Linux users at 210 million :-)

> Are we measuring computers or computer users? They won't be the same.

I'd say computers, not users. Machines are probably easier to count. Plus, these stats are mostly interesting to smack hardware vendors with so it's the hardware (sales) that matter.
number6x

Nov 30, 2006
6:13 AM EDT
hughsjr,

That is fantastic. I would think that would be a very accurate way of measuring active users.

Of course that could still be an undercount. Some installations 'may' mirror your updates internally, or do a standard build and then deploy that build on many machines internally.

It would be good to see similar stats from Red Hat, Novell, the major Debian Mirrors and others.

Thanks.
hkwint

Nov 30, 2006
6:39 AM EDT
So, that does mean, from the CentOS, combinind that with Distrowatch stats, we can make an estimate of the number of Red Hat / Ubuntu etc. users? This becomes interesting!
Sander_Marechal

Nov 30, 2006
3:51 PM EDT
hkwint: probably not. The distrowatch stats are very skewed. Based on this stat alone you'd yield 210 million Linux deskop users, or a market share between 20% and 50%. While that would be nice to dream about, it's way too much.

If we would be able to gather data like this off the top 10 Linux Desktop distro's then we could make a far more accurate guess.

Posting in this forum is limited to members of the group: [ForumMods, SITEADMINS, MEMBERS.]

Becoming a member of LXer is easy and free. Join Us!