Measuring the most Popular Linux Distros

[buster]

Anyone reading Distrowatch comments this week has seen the huge kafuffle about the "Major" distros. And this gets all jumbled up with popular, significant, upstream, historical.....

Anyway, I had my eyes opened today when I downloaded  cinnamon and xfce Mint isos. And by the way, the 19.1 has been around for awhile.

I always try to use a torrent when I can, and some distros work hard to get enough seeds and peers to make it work. But at the particular time I did this, late morning, there were well over 2000 people on the Mint download! So are there always this many getting Mint? Astonishing.

I don't know if this tells us whether it is 'major' or not, but it sure seems to be popular, no matter if MX has been getting more Distro hits lately. Seems like an interesting way to measure popularity.

[Jason]

Yeah, that could be a useful metric if compared to others.

The problem with a lots of counting is this, what are we counting? And what number is actually important?

To me, the most important number is how many people are actually using Linux as their primary OS - I care less about specific distros although that could be interesting, too. But back to the original question, what are we counting? To know how popular Linux is or a particular distro, we could count:

   
• Number of times that people read more about a distro by visiting its page (like on Distrowatch) - great to show what people are curious about but it doesn't show how many took the step of actually downloading it.
• How many actually downloaded it - better stat but how to measure? Torrents might be a good way to do this assuming every distro being measured has a torrent and it's leveled over time since it's going to spike right after a release. But people may download it and never get around to trying it out or they may try it for a few hours and hate it and it's gone.
• How many people actually use a distro - probably the best stat to really understand what is popular but it's also the hardest one to measure since their are potential privacy concerns (like tracking each user) - Fedora is interesting in doing this IP but as somebody pointed out, you could have one IP address with a router using NAT for a home/business network and lots of Linux boxes behind it. Also won't be able to effectively count those using VPNs.

Since most are probably using mirrors to download updates then perhaps the mirrors could log unique IP addresses accessing updates, be like what Fedora is using but able to compare multiple distros. I guess another way of tracking would be LUGs where they exist asking their members what they use and this information being aggregated across the world.