Sounds like a fun exercise, and 12 million does "feel" right. Some things I'd keep in mind - with the world economy in trouble, I wouldn't be surprised to see more people go for a free OS over the next few years. In addition, apparently Cuba, Russia, and China are trying to rid their countries of Windows in favor of Linux, so there may be a significant regional bias over the next few years.

The assumption that most of the computers are connected to the Web is wrong. If you go to any developing country the cost of the Internet connection doesn't make is possible for everyone to have one. I definitely say that the situation is improving.

Asa, 250 million * 5% / 65% doesn't give me 16.5 million for the "best case" - more like 19 million. So there is no real agreement at the high end.

But the resulting numbers do "feel" right, I don't think there are significantly more Linux PCs out there (with a strong emphasis on "PCs").

Have you considered comparing it to iPhone OS users to double check? It has a few benefits:

* Apple quarterly releases figures about the number of units sold
* >99% of all units will run iPhone OS
* When you come across iPhone OS in a statistic >99% of those will be on an iPhone
* Web statistics are likely rather reliable as almost all iPhone users will be regular web users due to unlimited usage plans

Does Mozilla even get update check requests when the OS handles the updating?

Now, looking at our Firefox numbers, we can positively identify Linux at at least 2% of total Firefox browsers based on update checks

Assumptions:

1. 90 - 95% of PCs are connected to the web regularly. There are far too many variables here and you making a wild guess.

2. 75 - 90% of Linux installs use Firefox as their primary connection to the internet, are you looking purely at Desktops? Do some Linux installs ever see a GUI? Again, I'm unconvinced that this is even a good guess.

3. 2% of Firefox's base is Linux. What are the basis of this statistic? Utilities such as Net Applications have ridiculously high error margins in the sub 2% share of the Internet so 2% of Firefox's share and we're talking probably error margins of atleast 1 order of magnitude.

4. The 3x multiplier is applicable to OS usage, not just browser usage, this is seems like a HUGE stretch.

5. The 3x multiplier is applicable across all OS', that is an average Windows user has the exact same browser habits as the average Linux user. This is even more laughable than the previous assumptions.

6. It's a "safe bet" that Linux users use the Internet on usage figures more than other OS'. I'm not convinced there aren't a lot of Linux geeks that set up media centre PCs or other offline PCs built generally for a single purpose.

You may or may not have a ball park figure, but I certainly wouldn't want to hazard a guess with all those assumptions.

Just a thought, but a lot of linux users mask their firefox to look like IE so they can get past pesky browser checks on some sites. I wonder if that would throw off some of these estimates.

Using the Firefox update checks to measure Linux users seems like it won't work.

For instance, I use Ubuntu (arguably the most commonly used desktop distribution these days, and in any case, it represents a non-trivial chunk of desktop Linux users). In Ubuntu, updates to Firefox are handled by the OS's package manager, and they even have disabled Firefox's built-in update checker. So, Mozilla will only see update checks from Ubuntu users if they have downloaded Mozilla's Linux build of Firefox and are not using the one provided by the OS (which I would say, most people would not do).

I wouldn't be surprised if other distros do something similar.

Al the comments above point modern linux distributions block automatic updates in firefox. Canonical's Mark Shuttleworts said they estimate about 8 million Ubuntu users, so I'd say though 10-16 would be a good guess.

As the comments above point modern linux distributions block automatic updates in firefox. Canonical's Mark Shuttleworts said they estimate about 8 million Ubuntu users, so I'd say though 10-16 would be a good guess.

dao and others, no, we do not see update checks for most Linux Firefox users. Sorry for the confusion. I wasn't looking at update checks, I was looking at plug-in blocklist checks which we do see, even for Linux distro-packaged Firefox users.

- A

This is rather a moving target complicated by duplication - I use a number of browsers besides FF - and differences in markets.
Brazil and Venezuela, for instance, can be counted on as those who have ditched Microsoft - as I have done myself.
While I'm aware my dropping XP because of being nailed by a Trojan puts me in a minority of violent reactions, the cellphone community is really where diversification is happening.
Rates in Asia for mobile net connection and connection speeds for PCs sound like we're a backwater. As a rural user pretty much locked into 1-2 meg speeds : and counting myself lucky to have had that for years : the frustration I feel while moving around without access and knowing antique tech of the phone companies is the main reason - this dialogue, while common, really seems to miss more interesting questions.

I'm one of them and proud of it!

(Though I may count as 2, since using Firefox and Shiretoko)

Firefox is indeed the major browser currently there is for Linux users, but due to Mozilla's prioritization, Linux bugs count keep rising more than other operating systems, only because Windows and Mac are P1, while Linux is P2. I'm getting complains daily from Linux Firefox users who see how Firefox behave on Linux comparing to Windows. Recent benchmark test even show how Firefox is faster under WINE than native on Linux (Slashdotted). Someone need to sit and start making Firefox better for Linux, because Linux users really like to use Firefox, and at the moment this barrier will fall, more users will migrate their Windows boxes over to Linux.

Mozilla should be more nice to the people who contribute to its success. As far as I can see it, Linux users are most of the people who actively promote and introduce Firefox to their Windows friends and colleagues. and it is shame that for them Firefox is less stable than for the average Windows user.

My two cents.

Tomer, Windows has a better compiler than Linux. That's a big part of the problem. Get GCC to produce more performant code, and Firefox will be faster on Linux native than under Wine. Also, on Windows, we do profile guided optimization when we build. Linux distros who are the main source of Linux Firefox don't do that. These are both issues that are outside of Mozilla's realm and if you're disappointed in that Wine to native comparison, you should chase down the Firefox maintainers at the various distros and get them to do PGO builds and you should chase down the GCC guys and get them to improve the compiler.

And, yes, I am blaming other people besides the Firefox developers for the results of that Firefox native vs. Wine test.

- A

manuthai

Just to make everything clear, I am using Mozilla Firefox binaries under Linux, and not distro-specific compilation, because I don't like the stuff they put inside my browser. Still, some things are broken, which make the Firefox usage horrible under Linux. For example, Gmail can slow down your browser when it loads, and Adoble Flash plugin will randomly crash your browser. If Linux users will prefer other browsers instead of Firefox (lets say Epiphany-gecko, as it is Gecko-based, but as Chrome will compile well under Linux, they might migrate to it), Mozilla will lose the major source of users promoting Firefox as a open-source browser.

Tomer, yeah, Mozilla doesn't do PGO builds either because "no one" uses our builds (except a few people like you) so it's not worth the effort on our end. The linux distros created a situation where there's no value in software vendors like Mozilla trying to make good builds for Linux because they lock most people in to their own packaging system. And even our own builds are also limited by GCC's inferiority to Microsoft's compiler.

Gmail doesn't slow me down at all in Firefox 3.0 or 3.1beta. Flash crashes are pretty rare for me and I mostly blame Adobe.

Also, I don't know what makes you think that other apps will do better under GCC than Firefox. Also, until Red Hat and Ubuntu move over to Chrome or Epiphany, I don't see Firefox losing many users on Linux.

Finally, I think you over-estimate the network effect of Linux users. Networking effects benefit from volume and the Windows user base and Mac user base of Firefox dwarf Linux. 240 million Windows and Mac Firefox users are doing a lot more Firefox promotion than 5-10 million Linux users. It's just basic math.

- A

Fuzzy indeed! In my home we have 9 PCs - 4 Windows desktops, 2 Windows laptops, a Linux dektop and a router and file server (Linux PCs).

Only one third of the Linux PCs I own are "counted" because two of them are just managing things around here. I think to your assumptions you can add there's probably close to a 1-1 ratio of total Windows PCs that are used to browse the web, etc. Because Linux is far more versatile, at least locally here, the ratio is 1-2.

You can set aside a Linux box for ANYTHING. You could do EVERYTHING on it, but if you had several to yourself you would probably not.
Linux is cheap enough (totally free if you get the right distribution the right way) that you can actually afford to install it twenty times on computers in your house (although I don't know anybody who does this).

Compare this to Windows, where the cost is so prohibitive that if you set a box up to do ONE thing, you are forced to set it up to do EVERY thing.
I have three Linux boxes, but I only use one to surf the net (except in emergencies). I could not afford three Windows boxes :-( and even if I could, I'm happier without them.

This comment is starting to sound like an attack on proprietary operating systems, on a troll to extoll the virtues of Linux - it wasn't intended that way, Windows is quite nice too even if it's not my taste. I don't mind if Windows users don't come to Linux ;-) as long as they're happy.

My point is, like the guy above me, only one of my three PCs is actually "counted" and it is the only one with Firefox installed. Oh, I use all three of them every day (although I don't sit down at any of them every day) but only one of them gets to surf the net. (and, dammit, I have to use dinosaur Windows computers at work)
When you do the numbers and come up with five, twelve, sixteen, twenty nine million users, keep in the back of your mind (no need to point this out, we know) that most Linux boxes might never get used to surf the internet. People who use it just seem to get a taste for using it in as many places as possible and for as many things (however obscure) as possible.

Asa, interesting to see someone from Mozilla addressing the TuxRadar "tests". What I thought was interesting was that testing using nightly builds shows that there's some discrepancy between different benchmarks.

I tested the latest 3.1 nightlies with a clean profile in Wine and Native Linux using Google's benchmark as well as SunSpider and Celtic Kane's test.

Interestingly though I got the same results as TuxRadar with Google's Benchmark, I found that SunSpider placed them almost equal and Celtic Kane's benchmark said that the native Linux version was almost twice as fast as the Wine version.

Why have you been wasting your time making Linux go twice as fast when we only represent such a small minority?! :P