Lies, damned lies...

Top edit: on re-reading this post it occurs to me that my intent is not entirely clear. I'm not attempting to show that the Distrowatch numbers are 'wrong' and the other numbers I post are 'right'. Of course, I don't believe that Mandriva is actually ten times as popular as Ubuntu, as the Softpedia numbers would suggest! I'm just trying to show that you can get wildly different numbers by looking at different places, which suggests that no-one's really nailed things down with authority yet. I think the gut feeling of anyone who's been around the Linux world in general for a long time on the relative popularity of distributions is likely to be more accurate than any single statistical snapshot.

Came across some interesting numbers today. Those of a statistical bent, when looking at Linux popularity, tend to wind up with the Distrowatch table, which doesn't make happy reading for a Mandriva person these days - Mandriva is 8th at 911 hits per day, well below Ubuntu and PCLOS at the top. The seven day figures (default is six months) are even uglier.

Of course, when we were on top of this chart for several years we always said it didn't really mean anything, so we can say the same thing now with a straight face: It doesn't really mean anything. It counts hits on Distrowatch's information page about a distribution, which is fine and dandy but really indicative of nothing more than how many people are going to Distrowatch for information about a distribution.

Let's see the other numbers. There's a guy who posts regularly to the Distrowatch Weekly comment threads called Peter. He looks at download numbers from the download server of a particular Australian ISP - Bigpond.

Australian ISPs have extremely tight bandwidth caps. Bigpond allows unmetered downloads for its users from its own server. This means there's a very strong incentive for Bigpond subscribers to download from that server (because the download will be fast and won't count towards their bandwidth cap). This makes it quite a useful test. By those numbers, Mandriva is seventh, with a note from Peter that it's moving up towards fifth or fourth. Ubuntu is first, and PCLOS is way down the list (he doesn't give it an exact place). In terms of numbers, Ubuntu has 1800 downloads, OpenSUSE in second has over 1000, then third is Fedora with around 800, and fourth Debian with around 550. Mandriva has around 450. PCLOS has 130.

Finally, there's Softpedia, a big software aggregation site. Take a look at the list of Linux distributions and sort it by Downloads, descending. Who do you see at #1? Mandriva, with over 180,000 downloads. #2, Kororaa (?!), at 160,000. Ubuntu Edgy is down there in 13th place, and PCLinuxOS is stuck in the middle of page 3 with 2,488 downloads. And no, they're not counting all Mandriva downloads since the beginning of time together - there's entries for older Mandriva editions throughout the list. (Though I think they may be combining 2007 and 2007 Spring).

All just goes to prove the old saying, as in the title. :)

I also wanted to update an old internal mail I did on the Most Pirated Linux Distributions (as taken from popular Bittorrent sites), but Demonoid is down right now, so it'll have to wait for another day...

Comments

brentbhola wrote on 2007-06-27 04:59:
Yes I had always wondered about the distrowatch numbers as well. Not that they are not very good releases, but I found it interesting that distros could rise so quickly or Mandriva could fall as fast as it did. I think it could just be search engine traffic bring people into various pages on distrowatch which inflates the numbers.
ofaurax wrote on 2007-06-27 08:02:
I think that the ISP numbers are what we search for. Distrowatch counts people that click on a link : how many of them will download/install it ?. For the most pirated distribution, this will not count freely available one (with a big green sticker "Download now !"). The ISP numbers (if we can have them all) would be very interesting : they count the effective downloads of a distribution, including current users who update.
yoho wrote on 2007-06-27 11:26:
Couldn't we have mirrors figures ? distrib-coffee ? free.fr ? they host other distro as well, so their numbers are probably interesting as well.
jkeller wrote on 2007-06-27 15:21:
Despite telling myself that the numbers didn't mean anything, I admit taking a certain amount of satisfaction that I'd chosen the "right" distro, back in the days of ye olde Mandrake's undisputed hold on the number one spot. I don't have that satisfaction any more, but then again I wasn't popular in high school, either (were any Linux users?). I can't imagine it changing my happiness with the distro, no matter which one. What I wish the armchair analysts (not directed at you, Adam, per your top-edit) would remember is that it's more like Google's zeitgeist: an indication of a distro's current interest levels. Big swings make a lot of sense in that context; today's Paris Hilton is tomorrow's Tonya Harding. There's a big difference between casual interest and commitment. Also, you don't get STDs from *any* Linux distro, so that's a plus.
adamw wrote on 2007-06-27 17:36:
yoho, I don't have those figures. I guess Nanar may have numbers for distrib-coffee.