I don’t know exactly what’s going on. But I do know that 3.5% to 10.5% in a few months is so beyond anything ever observed before, the notion that this is users adopting Linux over Windows/Mac cleanly and organically severely strains my credulity.
If you dive further into the stats, this increase is almost exclusively from the US. Mexico went from around 2 percent in May to 6 percent in June and July, and Canada actually went down from 5.6 percent in June to 3.8 percent in July, which, while it was larger than May’s 3.1 percent, seems like it’s within error.
Statcounter doesn’t have sophisticated bot detection like Cloudflare. Look at Cloudflare’s stats for the same period and filter for “likely human”. The percentage craters back to reality. So while I think “The loading times on headless alone would give them away.” is total bullshit, they can still be detectable using other heuristics and an awful lot of machine learning.
Statcounter monitors traffic on every website that implements it, the majority of those are likely Cloudflare or similar DNS, but we don’t really know for sure.
What does DNS have to do with anything? DNS isn’t involved in this data gathering. Cloudflare uses DNS to enable proxying, proxying is how they gather OS data.
Statcounter’s data is from visitors to pages with their JS embedded. That’s their whole business model. No one said it was visitors to the Statcounter website iself.
Have you considered that you might not actually know what you’re talking about?
You not knowing what a DNS is in this context or what CloudFlare does sort of makes a reasonable response impossible without typing multiple paragraphs.
My money’s on such a jump just being AI bots using headless browsers.
I don’t know exactly what’s going on. But I do know that 3.5% to 10.5% in a few months is so beyond anything ever observed before, the notion that this is users adopting Linux over Windows/Mac cleanly and organically severely strains my credulity.
If you dive further into the stats, this increase is almost exclusively from the US. Mexico went from around 2 percent in May to 6 percent in June and July, and Canada actually went down from 5.6 percent in June to 3.8 percent in July, which, while it was larger than May’s 3.1 percent, seems like it’s within error.
I did get kind of carried away creating so many LXCs and VMs in Proxmox between then and now…
Bots are rarely so subtle to be indistinguishable. The loading times on headless alone would give them away.
Statcounter doesn’t have sophisticated bot detection like Cloudflare. Look at Cloudflare’s stats for the same period and filter for “likely human”. The percentage craters back to reality. So while I think “The loading times on headless alone would give them away.” is total bullshit, they can still be detectable using other heuristics and an awful lot of machine learning.
StatCounter is hosted on Cloudflare DNS but if you’ve got more information about their data sources I could check if those filter bots as well.
StatCounter are their own data source.
Statcounter monitors traffic on every website that implements it, the majority of those are likely Cloudflare or similar DNS, but we don’t really know for sure.
The source is NOT people who visit Statcounter.
What does DNS have to do with anything? DNS isn’t involved in this data gathering. Cloudflare uses DNS to enable proxying, proxying is how they gather OS data.
Statcounter’s data is from visitors to pages with their JS embedded. That’s their whole business model. No one said it was visitors to the Statcounter website iself.
Have you considered that you might not actually know what you’re talking about?
You not knowing what a DNS is in this context or what CloudFlare does sort of makes a reasonable response impossible without typing multiple paragraphs.
I’ve been a paying Cloudflare customer for over 10 years and have had calls with their CTO. I don’t know what Cloudflare does? Lol.