Guest post by Internet News writes about a major mark for Fedora 10 release. Fedora remains the only distribution to publish it’s statistics and gathering methods openly and transparently. In any case, they reached 1 million active installations of Fedora Linux 10.
Red Hat’s Fedora Linux 10 has been out since the end of November 2008, and is now hovering around the 1 million installations mark. Fedora uses a system to measure active installations that check the update repositories in order to determine how many installations are in use.
When combined with other actively used Fedora distributions as of Feb 16, 2009, Fedora’s counting method reports 12,188,598 Fedora Linux installations across Fedora 7, 8, 9 and 10 releases. Obviously, some of those have been updated to later version already, and might have been counted twice.
On Fedora 10 in particular, in contrast with adoption for Fedora 9 for a similar period, Fedora 10 is at 115 percent adoption (that is a greater adoption rate for the first 12 weeks of release for Fedora 10 than Fedora 9).
I installed Fedora 10… Then removed it the next day and put Ubuntu back on that machine.
But, did they count that as an active install ?
How can they judge USE and TEST ?
Check the actual stats. They test EACH WEEK. You were counted one week, but not the next.
How did Canonical come up with its 8 million Ubuntu user claim back in 2006?
http://www.redherring.com/Home/20497
How did Canonical come up with the same 8 million Ubuntu user claim in September of 2008?
http://www.theregister.co.uk/2008/10/27/shuttleworth_ubuntu_commitm…
Noone from Canonical will go on record explaining how they come up with userbase estimates to the press which show absolutely no Ubuntu growth from 2006-2008.
You may dispute the Fedora numbers, you may even dispute the methodology…but point to another linux distribution who publishes a better methodology on how they come up with their usage numbers. If you can’t trust Fedora’s numbers, then you sure as hell don’t have enough information to trust Canonical’s claim.
Fedora cares about getting an accurate picture of the overall linux user install base. We want people to look at our methodology for counting users, we want best practises to be adopted by other distributions so we can get a very good comprehensive picture on the size and growth of the global linux userbase. That is why all the stuff we’ve done that goes into generating those statistics is openly available for reuse. We treat the client logs as sensitive data, but the processing scripts and the MirrorManager framework by which we collect client information is completely open for others to replicate and to integrate into their distributions.
You want your Ubuntu usage to count in a userbase size statistics that will hold up to scrutiny…encourage Canonical to publish a methodology and a set of trendable statistics by adapting Fedora’s MirrorManager service.
-jef
Fedora 10 is a great OS. Clean, solid, stable and up to date. Reminds me much on the experience I had with Windows 2000, the last good operating system from microsoft.
Edited 2009-02-25 23:51 UTC
Reminds me much on the experience I had with Windows 2000
I know what you mean. I felt the same way for a long time and loved 2K. I finally jumped to XP last year. Once you kill the unecessary services, add a decent firewall, and tweak the UI it runs as good for me as 2K. XP also runs media well on low memory devices like my eee pc.
I miss Fedora. I left it after the first version came out with Pulse Audio. I had so many media problems I got Ubuntu envy and caved into the popularity craze. I downloaded Fedora 10 Live to try it out but I am hooked on apt-get and the cool way Debian distros can be updated easily from major versions via a couple of terminal commands. I have fond memories of Fedora. I heard Pulse Audio is much better now and the system is very clean. Tough to go back once you get the flavor of Debian. Any others feel this way?
doesn’t ubuntu default to pulseaudio too?
Yes but when Ubuntu does it, it’s absolutely wonderful!
oh, my. it’s *usable* on 8.10, if all you need is basic sound support. on 8.04 it just flat-out didn’t work right.
many apps are still audio-broken in ubuntu’s 8.10 repos. not to mention, it’s still not acceptable for people using linux for pro-quality audio work, especially on ubuntu (and ubuntu studio). and last time i checked ubuntu still doesn’t make the pulseaudio jack module available. so, you really have to jump through hoops to get back to pre-pulse functionality if you have any special requirements.
i really have nothing against pulseaudio. but it certainly isn’t (yet) the perfect audio solution for everybody, even by the developer’s admission.
all that being said, fedora’s default pulseaudio setup has been much more problem-free (from my experience) than ubuntu’s.
blah.
Pulseaudio has been a nightmare for me. Sometimes, it took 90% off the processor and I had to kill it in order for the audio applications to work.
I removed that virus.. And I’m in heaven again.
Heh, Pulse has been both demon and savior, depending on the version I’m running and what hardware I’m running it on. On some hardware configurations I don’t need it, but when I start dealing with certain USB audio devices that do not do hardware samplerate conversion or do it badly, or internal cards that have this difficulty (ens 1370/1371 and various 8×0 cards), then Pulse is sometimes the only thing that even allows the audio system to function well. Without it, problems can range from artifacts in the audio to a nasty, ear-splitting hiss/screetch depending on what rates it attempts to mix together.
On the other hand, certain versions of Pulse have actually made it worse. Version 0.99.11, for instance, caused a nightmare with the soundcard I was using, causing the audio to only play for perhaps a quarter second before abruptly cutting off.
Pulse, I think, will eventually become a good solution. I do not think, however, that it was ready for general consumption when Fedora and Ubuntu started pushing it, nor do I believe it is at that level now. It still has quite a few quirks and a good deal of latency under certain tasks.
Pulse audio, also known as “Let’s re-invent ESD again, because it worked so well the last five times,” is truly horrifying. Rather than solve the Linux audio problems correctly someone decided it would be a good idea to write a user space management program with all of the problems of the old management programs, only it’s better now honest.
Some of the /theory/ in pulseaudio sounds great and should be in the kernel. The implementation suffers from acute “Doesn’t Work” syndrome and has the lovely effect of making something simple really complex. It’s nice that someone cares about audio, but it would be nicer if they wouldn’t introduce yet-another universal audio API. If there’s something wrong with alsa, fix alsa.
If there’s nothing wrong with alsa, why are we using pulseaudio?
Fedora, Ubuntu, OpenSUSE, Mandriva and many others use PulseAudio by default and do it for good reasons. ESD and PulseAudio are both sound servers. Beyond that they are not really comparable. One is dead and broken. Other is actively maintained and has a lot more features. ALSA is too low level to do a number of features that PulseAudio does and that competing systems like Vista and OS X have done for a long time as well. Handling of bluetooth headphones or per application volume control for example.
A. ALSA depends on hardware mixing and not all sound cards are capable of reliably mixing different sources.
B. ALSA cannot (and should not) be able to control the volume level of each source stream. (Try listening to a Internet radio while playing a -loud- computer game)
C. Pulse is network transparent.
etc.
I don’t use pulse on most of my Fedora machines (I tend to disable it) – but I have no doubt that I -will- use it once it matures.
– Gilboa
But that’s the thing… WHEN it matures. As of now, it shouldn’t even be used as the default sound engine because it crashes and slows down the comp. why release an LTS version of Ubuntu with an important product still is still in Beta or Alpha?
Why did Canonical decide to do that? You got to ask them. Nothing to do with Fedora.
yum remove pulseaudio
you dont actually NEED it to play sound.
Not really. I don’t revere apt and dpkg as, it seems, most on here do. In fact, I’m heartily sick of it to be honest. It’s great when you don’t have any troubles, but when you do they can be annoyingly difficult to trace. My favorite is when you simply get:
E: sub-process dpkg exited with status 1.
And no other details are forthcoming. Honestly, did they take lessons from Microsoft when it comes to error messages? That was less detailed than any error even Windows gave me. Obviously, status 1 means some error caused dpkg to die, but many times you don’t actually get to see what that error is and have to hunt back in a log to find it, and that’s when you’re lucky and the error gets logged. That, plus any error on the server-side db corrupts your client side db as well due to how closely linked they are, and if you think you can have dependency hell with RPM, you don’t even know what can hit you with apt. Granted, it hits you less often, but it’s a real bitch when it does. It can make dll hell in win 95 look like child’s play by comparison.
The whole bloody thing is a mess, in my opinion. Rpm and yum have their drawbacks, but sticky situations are much easier to resolve with them. Each package header is separate, so if there’s an error in one it doesn’t blow your entire yum database to hell and, if a broken package forces you to override a dependency or, as can also happen, you know you don’t need a particular dependency, yum/rpm doesn’t give you crap about it from there on out. Apt will constantly annoy you and try to fix what it considers a broken package. I’m in charge of the system with yum, with apt I feel I have to constantly battle it for control.
Of course, given my preference, I’ll take pacman over dpkg or rpm any day. Now, that’s a nice package manager, pacman.
Fedora can be updated with a couple simple commands the same way. We use a “preupgrade” package to download the content you need, while you keep working on whatever you like. Then you reboot, the installation finishes, and you’re running the new version.
Most people who think there’s some wide gulf between apt and yum functionality haven’t used one or the other in a long time.
And they say Linux hardly has any marketshare …
And they say this because it doesn’t – how does this surprise you?
Fedora, one of the ~500 Linux distributions reports 12M users. Ubuntu has similar numbers. Netbook sales indicate 30% Linux. Some estimates now put Linux installed base at double the OS X installed base. Microsoft has declared their number one competitor to be Linux, not Mac.
What are you looking at if you think Linux market penetration is insignificant? Market share? That’s the number of copies of the OS sold. Since Linux is free, it doesn’t apply. The numbers you want to look at are the installed base – the number actually in use. Unfortunately, just as with Windows, this can only be estimated. For example, I have a dozen Windows licenses and a couple of OS X licenses, but all machines are really running Linux exclusively (the installed base).
If we could get hard numbers from Xandros/Asus and Linpus/Acer concerning the number of pre-installed linux netbooks that would go a long way to addressing traditional “marketshare” in the sold in a box sense.
Those vendors account for something like 70% percent of all the netbooks sold last year, but we don’t know what the split is between linux and XP models.
The userbase estimates for self-installed linux instances are indeed harder. And Fedora is trying very hard to carve out a methodology for other linux distribution to re-use so we can work together to get a global picture across all distributions of what our userbase looks like. There’s no secret sauce in how Fedora count users.
-jef
“What are you looking at if you think Linux market penetration is insignificant?”
Sure. I was really only commenting on the comment, well, I commented on, which in turn seemed to find 1 million impressive. However, I am not sure whether that was meant to be ironic or not.
OK, I don’t know about 12M Linux installs, but I’ll go with that.
So then: if you search for PC sales world-wide numbers, I found 75M for Q4 of 2007, I believe (rounded of for convenience here). Also for convenience, lets assume that is kinda representative and I’ll generously round this of by a quarter billion in your favour to make it 1 billion PC sales in the past 4 years (past 16 quarters). Of course, there are vastly more than a billion PCs around.
Now, don’t blame me for the maths if I tell you that 12 million out of a billion (which isn’t really a billion in the first place) is 1%. But since it is more likely that there are +2 billion PCs actively used out there, that puts your 12 million linux installs at 0.5%. Can you confirm that you agree with that?
Also, who cares about 500 Linux distris out there? Half of them only show up on distrowatch because I made the single download that caused it to be listed there in the first place = forget about most of the ditros, they carry no weight.
0.5-1% is irrelevant, can we agree on that? I have Suse, Ubuntu, PCBSD and Opensolaris here — that doesn’t mean everybody else has, too.
No.
An AT Internet Institute (formerly Xitimonitor) survey gives Linux a 1.2 % share of web users.
http://www.atinternet-institute.com/en-us/internet-users-equipment/…
I don’t know who owns AT Internet Institute or whether they have a marketing agenda of their own. The numbers are based on a very large sample of web traffic, but I don’t know if they disclose the sites that are monitored. I think it would be easy to get a bias even in a very large sample by selecting certain types of sites.
Anyway, 1.2 % of web users is tens of millions of users, and it is clearly the most common OS after Windows and Mac OS X. It’s not irrelevant by any means, it serves as a demonstration to hardware manufacturers that Linux is a viable and low cost OS. The manufacturers are finally taking notice and starting to offer pre-installed Linux systems.
Critics have said that Linux is a toy, irrelevant, about to be overrun by the next version of Windows etc. etc. for as long as Linux has existed. The installed base of Linux has grown steadily all the while, albeit slowly on the desktop. This is how free software on the desktop tends to work: there is no great redesign followed by a marketing push. Instead, different parts of the technology get better all the time and while it may not look like much at any time, it adds up. The parts are now falling into place at an accelerating speed.
I am running Fedora10 and long time fan of Red Hat, however the advancements in the Linux distro arena are breathtaking and very difficult to keep up with.
I am amazed at how may improvements and new features they include and they work extremely well.
Network Manager has to be the best setup for Laptops or other devices that get switched from wired/wireless.
Plus, SELinux has come a long way as well, I always leave it on and after learning how it works with the file system it makes logical sense.
Overall happy Fedora user.
It’s not bad for that. However having a computer connected to both a wireless and a wired connection is a pain. I didn’t really follow this, but some time ago it was not possible to use DHCP on both connections at the same time.
This has not been a problem for some time. I have a setup at home with an internal DHCP server and a laptop that’s usually wired and wireless, and I have no problems getting DHCP addresses on both. One only uses one at a time, of course. When I unplug the wired connection, I’m immediately on my wireless connection, and when I plug back in, I’m back on wired. When I do this, of course my default routing changes.
NetworkManager is nice compared to the horribly-complex network config file setup on RedHat-based OSes. However, compared to the beauty and simplicity of FreeBSD’s rc.conf or Debian’s interfaces file, NetworkManager is more of a hindrance than a helper. Especially since it keeps it’s own, completely separate configuration databases. Why, of why, oh why, do all these new GUI tools have to be replacements for the text config files … and not just new ways to configure the text files?
NetworkManager does automatic switching between wireless and wired networks which is critical for any laptop in addition to lot many things like network sharing. This functionality has really nothing to do with configuration format differences and that is the reason why all the mainstream distributions are now including it by default. NM does use the old configuration files when it can but much of the information it needs doesn’t exist in any configuration file at all. You need to look at how it works a bit more, I think.
A properly configured interfaces(5) file will also do this. Same with a properly configured rc.conf(5) and wpa_supplicant.conf(5). And it’s not rocket science to configure either of them.
You still have to fiddle everytime you switch networks with the configuration files and that’s the reason Debian includes NetworkManager like every other mainstream distribution.
Yes!
Machine-editing interfaces(8) on Debian is quite easy. It integrates into most networking stuff quite well, even wireless.
On my Debian box I tried networkmanager and was constantly frustrated. My manual setup of interfaces is much more reliable and took little time, even though I did it by hand. If I had a wish it would be for better integration between interfaces and wpa_supplicant, which btw is really sexy.
running on 3 of my PCs. Ubuntu 8.10 @work o Vmware
The article sounded surprised that Fedora 8 still accounted for the most active users. I think a lot of the blame can go to Fedora 9, which was one of the buggiest Red Hat/Fedora releases of all time and even had the unbelievable gall to ship a beta version of the X server for 4.5 months of its lifetime.
This meant that there was no ATI 3D driver for Fedora 9 for 4.5 months and Fedora 10 had a not dissimilar problem for about 2 months (until, eventually, RPMfusion put it into its testing repos). It’s why I stayed on Fedora 8 for over a year (an all-time record for any Linux distro for me) and waited until January 2009 to finally switch my primary desktop to Fedora 10.
BTW, I’ve removed NetworkManager from Fedora 10 because I don’t like it starting networking up after I login and wiping my custom resolv.conf…
How is Fedora the only distribution to “publish it’s statistics and gathering methods openly”? What could be more open than the way Debian does it with popcon?
http://popcon.debian.org/
Ignorance is no excuse for such bold claims.
Yes, ignorance is no excuse which is why you should look at
http://fedoraproject.org/wiki/Statistics
No other distribution does this. Popcon is not the same thing at all though useful in other aspects.
Other than per-IP information, which I admit is useful and which popcon does not provide, what is it you claim Fedora provides that popcon doesn’t? Edits to a wiki? Hits to their web site? I don’t think those are very meaningful beyond a basic “Yes, people are interested in us” level. Publishing mailing list traffic details would be about as helpful to determining user base.
The original statement was that no other distribution
1) Publishes its statistics (with the implication of ‘on install base’)
AND
2) Publishes its gathering methods
Openly and transparently.
Well, popcon does do those things. Popcon’s statistic for the base-files package is effectively an opt-in Debian user base measurement. You can argue over whether its data is more useful than Fedora’s, but the statement is still false.
Popcon’s failing is that it is voluntary and off by default, so the numbers are certainly low. This does not make the methods used to obtain them hidden, nor does it make the raw numbers secret.
You still missed out something. Smolt system profiler has over one million unique system stats presented at
http://smolts.org/static/stats/stats.html
The combination of all these stats presents a good overall picture and better than any other distribution is unique to Fedora. Some other distributions might do bits and pieces but nothing really comparable.
I will reiterate: Whether it compares or not is not relevant. If other distributions collect statistics on usage, publish the statistic collection method publicly and provide the raw numbers then the statement in the summary is false.
All I got to say, is to read things in context.