Comment by Ologn | Hacker News Reader

Comment by Ologn | Hacker News Reader

Parent Full thread

As Wikipedia has an Alexa rank of 6, and is the most popular web site I know of that gives an occasional survey of its server logs, I tend to look at it for information on web clients. Like their October 2012 report ( http://stats.wikimedia.org/wikimedia/squids/SquidReportOpera... )

For Linux client systems that use Wikipedia, Linux kernels with Android dominate. There's about 7 Android clients for every 1 Ubuntu client. There's about 45 Ubuntu clients for every Fedora and SuSE client. There's 2 Fedora users for every Debian user. And so on - less used Linux clients are listed on the report as well.

Netcraft used to have a decent survey of which web servers were running Linux, but I don't see any recent reports from them on that. With the rise of EC2 and PAAS/IAAS I would guess the Amazon Linux AMI would be one of the new VPSs/instances/"servers" that is around more nowadays.

A lot of Linux distros don't expose the distro name to the browser these days, for security reasons. Ubuntu still does, though. That's why "Linux Other" is so high now (double Ubuntu) when it wasn't a couple years ago http://stats.wikimedia.org/archive/squid_reports/2011-03/Squ... .

Agree. Probably the only valuable statistic here would be of user preferences. Again, it reflects the biases that reddit users might have. But it is likely that they are more techsavy than the average population and likely to be an opinion leader among their peers.

Do Linux based spiders and crawlers count? If so, has that skewed the results in a meaningful way?

Unless the web crawlers are intentionally spoofing their user agent to identify as browsers, I don't think that they are counted.

Whether they count or not, I would expect to see "regular" users to be orders of magnitude more common than web crawlers.

But web crawlers generate orders of magnitude more traffic than regular users.

True, but if you're tracking users, you'll normalize for that (tracking cookies, bot exclusion, IP, etc.). It's somewhat nontrivial, but most web analytics packages will give you a decent first cut.

Reply on news.ycombinator.com