Slashdot Mirror

I don't know, but it is pretty much the only kind of mainstream OS which has problems with handling IPv6 properly:

http://fud.no/ipv6/

Let me correct myself; according to these slides, the initial brokenness per the end of 2009 was at 0.2%, which means the situation has improved an order of magnitude in terms of affected users.

An experiment done in 2010 on vg.no, a large news site in Norway, showed that up to 0.06% of their users could not connect to a dual-stack host, largely due to configuration problems in Mac OS X and Opera.

The problem occurs when the computer mistakenly believes it has a working IPv6 connection, so the browser tries the AAAA record first, which either outright fails, or causes a time-out delay before it falls back to IPv4.

These problems were fixed in newer versions, and the client loss is now at around 0.015%.

That's pretty low, and it was enough for vg.no to enable dual stack for the main website, just as these sites are now doing for a day.

Still, even at 0.015% that works out to quite a lot of Google and Facebook users.

(Source: http://www.fud.no/ipv6/)

One of the largest Norwegian newspapers have been running an automated test in an iframe - it was mentioned in an Slashdot story last year.
They were measuring connection problems for around 0.02% of their users this December, right before they enabled IPv6 for their main site.
Graphs and more info.

[...] if a website advertises itself as simultaneously IPv4/IPv6 compliant, and someone's computer/browser thinks they are IPv6 compliant but their attempts to connect via IPv6 don't make it through (ISP? router? modem? who knows), their connection times out and the site is unreachable.

More precisely: if the DNS has both v6 (AAAA) and v4 (A) records for the site's name, and the client prefers v6 connectivity over v4, and a v6 connection can't be established for some reason, the site will appear to be broken. Most large sites have measured this kind of brokenness, but haven't published their methodology nor results; there is an exception, but it's limited to Scandinavian users. It is nevertheless a very interesting analysis, which basically suggests that eliminating just two sources of brokenness (OS X < 10.6.5 and Opera < 10.50) would practically eliminate client loss.

What you didn't notice the link to The Consulant's Report? FUD? No.

...when, most notably, also Opera versions prior to 10.50 are affected!

OK, OK, and apparently many Linux distros prior to recent patch of glibc. From the original source (love the url): "Also I'd like to thank Opera Software for working with me and fixing the problem in their browser, and Fedora, Canonical, Gentoo, Novell, Mandriva, and Debian for applying my patches to glibc in their respective Linux distributions."