How many Fedora users are there? [LWN.net]

The Fedora Core 6 distribution is nearing release. Even after the recently announced delay, the final version of FC6 is expected to hit the net on October 17. So one would assume that there would be little call for controversial changes at this time in the cycle; the Fedora folks would be expected to be concerned with fixing the final problems and getting the release out the door. So the documentation group was a little surprised, at the end of September, when a request to modify the Firefox startup page showed up.

In particular, the Fedora leadership wanted that page to include a tracking image - an image hosted on a Fedora web site which would allow the project to track how many people were starting up Fedora's version of Firefox, and which IP addresses they came from. It would appear that few people had any sense that there might be objections to this technique; the resulting discussion seemed to take them by surprise. But a discussion did result, focusing on a few questions: why does Fedora want to track its users, why the hurry to get this change into FC6, and isn't there a better way?

At the moment, it seems, the Fedora Project has very little idea of how widely used their work is. That is an ignorance they share with a great many free software projects, but Fedora's situation, it seems, has the potential to make that ignorance expensive. The best description of Fedora's motives came from Greg DeKoenigsberg; it is worth quoting at length:

Really, this question should be asked this way: "are metrics so important that you're ready to risk alienating some users and contributors to get them?" And the answer to that question, from my perspective, is "yes".

Why? Because, like it or not, every funding conversation inside of Red Hat's walls begins and ends with metrics. If it isn't measurable, it doesn't exist. Fact.

This is especially important in the case of Fedora, because Fedora doesn't make any money directly for Red Hat. We continue to develop Fedora because it serves other purposes. Research and development. Quality Assurance for RHEL. The ethics of continuing to provide free software, which is important to all of us. And, most importantly from my own perspective, *community mindshare*.

If we can't quantify Fedora's mindshare in some way, we lose one of the *major* rationales for making the Fedora Project stronger and more independent. Every time a Red Hat executive asks "how many Fedora users are out there?" and we answer "oh, somewhere between 100k and a few million," we make it *that* much more difficult to defend Fedora from bad Red Hat decisions. If a Red Hat executive has to choose between giving resources to RHEL and giving resources to Fedora, and if he's got dollar figures on one side of the ledger and hand-wavy "mindshare" guesses on the other side of the ledger, he's going to choose RHEL. Every single time. I've seen it happen, again and again and again and again. And again.

Fedora has, slowly over the years, become a more open and transparent free software project. It is also clearly a successful project, with a large (if unknown) number of users worldwide. But the fact remains that Fedora is a Red Hat project, with Red Hat being the source of almost all of the funding that keeps Fedora going. This funding is a generous gift from Red Hat to the community (though Red Hat certainly benefits from it as well), but it puts Fedora into a strongly dependent position. Fedora must keep Red Hat happy, and convince Red Hat of its importance, if it is to continue to be funded properly.

According to Max Spevack, there is no concern about Fedora funding being cut; this exercise is, instead, about getting that funding increased. But the evident level of concern belies that claim somewhat. Even if there is no discussion of cutting Fedora funding now, it seems like a subject which could come up in the future. Red Hat is becoming just another company in many ways, and it will make the calculations that companies need to make to survive. It would not take too many bad quarters for Red Hat to start looking very hard at the money spent on Fedora; managers under pressure to improve their numbers can be very short-sighted at times. So it makes sense for the Fedora project to be concerned about its ongoing relationship with Red Hat.

It almost seems that something must have happened to reinforce this idea in the minds of the Fedora leadership. If so, they aren't talking about it. But they have decided that it is important to get some sort of mechanism into FC6 which would give them at least rudimentary statistics. Waiting another cycle for FC7, it seems is not an option. Given the short time available to put anything into FC6, the Fedora folks settled quickly on something which would be easy to implement: a tracking image.

There are obvious problems with the tracking image idea, starting with the privacy concerns. Not everybody wants to be tracked in this way. People with this sort of concern may also not be much comforted by the Fedora privacy policy page, which leads off with this text:

THIS IS A DRAFT. It may not represent the final document, and should not be used for anything other than informational purposes.

Beyond that, it has been pointed out that this technique only yields IP addresses, which will only be correlated with the number of actual installations in a very rough manner. But that information, it seems, is much better than nothing.

There are alternatives. One idea which has been discussed is a brief user survey which shows up at the end of the installation process. Users could then provide some information - or, crucially, choose not to. Nobody seems to think that such a mechanism could be added to FC6 at this late date, however; though it could show up in FC7.

The Fedora folks could also take advantage of the fact that a new Fedora installation already phones home. It is all for the best of purposes: the yum-updatesd daemon, which runs by default, goes to the central Fedora server to download the lists of repository mirrors. The project has not been using the tracks that this activity leaves - but they could. Greg describes it as "an absolute no-brainer":

The rich irony here, of course, is that rather than tell users we're tracking them, we will instead be able to track them invisibly through the normal operation of their systems. But I'm perfectly happy either way, so.

This approach is not perfect either. It fails on systems which are offline, while every system running Firefox has a high probability of being connected. It also cannot distinguish systems which are likely to be "desktop" systems - information which is apparently of interest. But it's there now and, as Greg points out, it doesn't seem to set off alarms the way a tracking image would. Hopefully Fedora will share the conclusions it draws from this data - and make good use of it to convince Red Hat management of the project's importance.

The LWN site is currently under high scraper load, so comment display has been suppressed for anonymous users. If you are a human, you may read the comments by clicking the button below:

Note: you can avoid this step in the future by logging into your LWN account.