Fedora Metrics Help Whole Linux Community

lisah writes "When Fedora released Fedora Core 6 late last year, the team decided to track the number of users with unique IP addresses who connected to yum in search of updates for a new installation of FC6. According to the data they collected, FC6 crossed the one-million user mark in just 74 days. Fedora Project Leader Max Spevack says that while it's great to use metrics to better understand what users want, the real value lies in its ability to encourage hardware vendors to more offer more Linux-oriented goods and services. Spevack told Linux.com: '[W]e always say we wish hardware vendors had more [Linux-capable] drivers. Well, if you can go to them and say, "Hey, there's millions of people using this," then maybe they will listen. In the real world, you need data to prove your case. Well, here it is.'" Linux.com and Slashdot are both owned by OSTG.

Re:Saddly...

[spevack]

Actually, the Fedora folks address that very point. Quoting from the Fedora Project wiki, and it's page on Statistics:

"Accuracy of metrics

We believe it is reasonable to equate a "new IP address checking in" with "a new installation of FC6", with the following caveats:

1. Users who have dynamic IP addresses will likely be counted multiple times, which inflates the number by some amount.

2. Users who are behind NAT, corporate proxies, or who rsync updates to a local mirror before updating will not be counted at all.

The anecdotal evidence that we receive from different groups, companies, and organizations makes it quite clear that group (2) is significantly larger than group (1). As such, we believe that the true numbers in the field are higher than the numbers on this page."

Impossible to install without connecting

[currivan]

I just installed FC6 on a machine yesterday, and they made it impossible to do anything without connecting to their server. I'm keeping the machine off the network, but apparently there's no way to install packages from the DVD without first downloading the update lists from their mirrors.

The Add/Remove gui (and yum) crashes if DNS isn't available. After some research, I was able to hack the yum .repo files to point to the DVD instead of the internet, but it still crashes with mysterious errors about media uris. I finally gave up and installed Ubuntu instead. So no, this doesn't help the whole Linux community. We'd be furious is Microsoft imposed this sort of requirement on new installations.

RH response to Ubuntu's 8 million number?

[Locutus]

Given the numbers coming out, I'd think that it sure can't hurt for these guys to post the number they are.

Here(2nd page ) Mark Shuttleworth mentioned Ubuntu having 8 million active users:

http://redherring.com/PrintArticle.aspx?a=20497&se ctor=Briefings [redherring.com]

Now what are the hardware vendors waiting for? Permission from Microsoft?

LoB

Re:Impossible to install without connecting

[unchiujar]

What about yum --disablerepo=* localinstall or rpm ?

best effort + transparency

[spevack]

I'm the guy who actually maintains that Statistics page on the Fedora wiki.

The real "story" here is a couple of things:

THING 1 -- We're making the best effort that we can at showing the world how many installations of Fedora Core 6 we know about.

THING 2 -- We're being upfront about the assumptions and caveats that go along with that number. Quoting:

"Accuracy of metrics

We believe it is reasonable to equate a "new IP address checking in" with "a new installation of FC6", with the following caveats:

1. Users who have dynamic IP addresses will likely be counted multiple times, which inflates the number by some amount.
2. Users who are behind NAT, corporate proxies, or who rsync updates to a local mirror before updating will not be counted at all.

The anecdotal evidence that we receive from different groups, companies, and organizations makes it quite clear that group (2) is significantly larger than group (1). As such, we believe that the true numbers in the field are higher than the numbers on this page."

THING 3 -- We're also being upfront about how that number is generated.

I'm not trying to spin the data in any way. I'm just putting it up there, and trying to do so as objectively as possible. Anyone can draw their own conclusions, or compare it to data from other distributions, if you can find similar reporting.

Re:Saddly...

[Lord_Slepnir]

The numbers will be inflated, but also deflated by places like the one where I work that have multiple FC6 hosts behind the same router.

Re:People != Computers

[RyoShin]

You are quite correct; one person would not download the CD three times.

However, that's not how they're collecting the data:

the team decided to track the number of users with unique IP addresses who connected to yum in search of updates

While you need only one CD to do multiple installs, it is my understanding that each machine has to run YUM itself. They've also thought of what you mentioned.

According to Spevack, it's not enough to simply count how many times the distribution has been downloaded

Now, the article does go on to say

Cacti tracks the number of unique IP addresses that connect to yum with a new installation of FC6

So, if YUM/the tracking software can differentiate between a fresh install and a regular update, then the number of connections better correlates to new users.

Re:RH response to Ubuntu's 8 million number?

[spevack]

The key difference, IMHO, is that in Fedora we are trying to demonstrate *where* our numbers are coming from, as opposed to just giving a number with no context.

It's also important to realize that this metric is just for Fedora Core 6, not "all instances of Fedora 1-6".

Re:1,000,001 I can't switch but would like to

[Intron]

So just fire up a live CD with a recent kernel and try it out. You don't have to upgrade if it doesn't work. Hardware drivers are in the kernel, so just testing the right kernel on your system will tell you whether it works (mostly).

FC3 uses kernel 2.6.9
FC6 uses kernel 2.6.18

Re:1,000,001 I can't switch but would like to

[FudRucker]

you need to learn to use Slackware, it is the best distro for old hardware...

Re:1,000,001 I can't switch but would like to

[Znork]

Personally, I rsync from a mirror and have a local repository, so I have a whole bunch of machines that dont get counted. Stuff like that will result in the numbers being a bit off.

"so I'm too afraid to switch from Core 3 to 6."

If you upgrade that rarely, I'd suggest you take a look at CentOS. CentOS 4 will be a far smaller leap (RHEL4 is close to FC3/FC4), and you'd be on a maintained platform again.

Re:Impossible to install without connecting

[LnxAddct]

I just installed FC6 on my macbook pro over the weekend, and I had no internet connection at all during the entire process (I regularly work offline). It worked fine, so I can only assume that your case is an isolated incident.
Regards,
Steve

Re:But....

[spevack]

the day they start requiring registration or creating GUIDs is the day I give the shove to Fedora -- and I've been and RH users for 8-9 years.

As the "Fedora Project Leader", the Fedora buck stops with me, so to speak.

And I promise you that I will NEVER require anyone to "register" Fedora in order to download updates, or stuff like that.

Neither I, nor the Fedora Board, which is Fedora's governing body, will allow some sort of "required registration" in order to get the full Fedora experience.

Download. Install. Update. If that's the extent of a person's interaction with Fedora, fine by me. We hope, of course, that there will be a fourth step, that being: Contribute

Re:Impossible to install without connecting

[jon_burgess]

Unfortunately the repodata provided on the CD & DVD is not useable by yum but creating a local yum repository [city-fan.org] is quite easy once you know how.

Installing packages from the original media is great just after you've loaded the system, but remember the good old days when you would be given the a prompt like: to complete this change you need to insert disk 3 of the installation media. Good luck finding the original disks a year or two after installing the PC.

I believe the majority people are happy that yum is preconfigured to download and install the most up to date version of a package from the internet whenever they need to install something new.

Re:best effort + transparency

[spevack]

Well, we'll keep doing what we currently are doing. In addition, the idea currently under consideration is an OPTIONAL screen in firstboot where a user can choose to let us know more about their hardware and/or installed package set.

KEY POINT TO MAKE: If a user says "no, go away and leave me alone", we will respect that.

To anyone who wants to be part of the discussion, feel free to follow the Fedora Infrastructure list.

http://www.redhat.com/mailman/listinfo/fedora-infr astructure-list [redhat.com]

Huh?

[EXTomar]

I just did a retro-fit upgrade and an install on two machines and neither went to the "yum" repository mirrors to do an update till after they finished their first reboot where I had to activate the update manually (and get the gpg keys installed).

- I remember that "install" at some point gave me an option to install against latest package in the "yum" repositories, which I do not do for speed.
- I remember the "upgrade" and "install" screens from Anaconda being different. The "upgrade" never asked me to update against the "yum" repositories.

"pup", which is the graphical tool analog to "yum", handles rotating through the mirrors properly as far as I remember where it just fails over to the next if the current one can't be reached. I've had my Internet die while trying to do this, I don't recall it ever crashing on me and this is doing many installs and upgrades across every version of Fedora.

I don't blame you for switching to something else given these problems. I'm just stumped how you got these problems.

Re:NAT is even more significant.

[Karzz1]

From the article:
"We believe it is reasonable to equate a "new IP address checking in" with "a new installation of FC6", with the following caveats:
1. Users who have dynamic IP addresses will likely be counted multiple times, which inflates the number by some amount.
2. Users who are behind NAT, corporate proxies, or who rsync updates to a local mirror before updating will not be counted at all.

The anecdotal evidence that we receive from different groups, companies, and organizations makes it quite clear that group (2) is significantly larger than group (1). As such, we believe that the true numbers in the field are higher than the numbers on this page. "