When I spoke to Google after W6D last year about the brokenness from JANET, I was told that bad 6to4 clients were certainly a significant factor. The failure rate from AS786 to dual-stack destinations vs IPv4-only was 4% or so, which was way too high for their liking. If you removed Mac clients the failures dropped an order of magnitude.
Back then they were seeing 1.33% of traffic from AS786 able to do IPv6 natively.
But that was a year ago. And Mac OSX now behaves rather differently.
Sites should check how much 6to4 traffic they have. The thing is now, with rogue RAs from Windows ICS issuing 6to4 RAs, hosts talking to Google etc in those networks (probably campus wireless networks) may use 6to4 rather than IPv4. If they do, and 6to4 fails, that will raise the brokenness stats.
The other thing different to last year is the wider implementation of happy eyeballs... now in some form in Firefox and Chrome, which will try IPv6 and IPv4 almost simultaneously, rather than waiting to fall back having tried IPv6 first.
Interesting though!
Tim
|