Linux enthusiasts rejoice! After a long journey, according to StatCounter’s data, by June 2023, Linux has achieved a 3% desktop market share. Learn more about it here!
@linuxiac I don't see an explanation of how this information was gathered.
Most distros don't run telemetry and download of an ISO isn't evidence of an install. I certainly keep ISOs local and install from there so likely most Linux users aren't counted.
Browser reporting which is very easily spoofed? And has to be sometimes because idiots write stuff for Chrome like they used to for IE?
Hows that line go? "...everything is made up and the points don't matter."
@melroy@linuxiac I don't buy the accuracy at all. Between functionality reasons to modify the string, and measures browsers, and more generally users of Linux, take to obtain and maintain privacy there's no way this info is even reasonably accurate.
I'd also argue this doesn't capture things like Steam Decks where users are using the machine daily to game and aren't even opening a browser.
So many other activities on the nets, and off, that aren't just web browsing.
OK fair enough. But then again the statistics was already containing incorrect data from day 1. So when we compare relative numbers from 5 years ago with now, it should give us some inside. Right?
@melroy@linuxiac on its face that's sound logic. However 5 years ago GDPR was just coming into force nor have major steps the likes of apple and the browsers have taken to protect privacy. I'd say given the largest constituents in this data are the least caring of the software in regard to user protections this data is probably further skewed. Entirely IMO of course I don't have data. But then that's the catch, how do you prove a negative? You just can't.
@Tekchip@melroy@linuxiac I am not sure there is any need to prove a negative. All data sets will be incomplete, but we can still use statistics to extrapolate. Otherwise we are just arguing 'how do we know anything'.
According to their FAQ they source from over a million sites and are geographically diverse, so these numbers are within any reasonable tolerance.
@keithnator3000@melroy@linuxiac that's a massive assumption when the asertion is that you can't meaningfully measure something.
The same reasons I outlined that Linux cannot be meaningfully counted, by it's nature, the nature of it's users, and ease of agent modification, scales equally regardless of the data set size or geographic coverage.
You should strive to know of course. But you should also strive to accept what you don't know rather than make assumptions.
I mean what assumptions are missing? The counter looks for desktop users.
If there was a massive under count due to agent mod then we might also be seeing them more IRL. I think their figures match field observations.
I don't dispute that a linux user is more likely to change theirs, but I don't think it might change the 3% figure all that much.
Getting numbers on user agent mod extension downloads and comparing that to the 3% number could be interesting to try and estimate how many extras might be missing.
'Meaningfully measure' is also a loose term. Statistics is at its core about not being able to measure or count things in their totality. I don't think missing not at random applies here.