[AusNOG] Stupid web browser behaviour Was: My Predictions for the ISP Industry

[Geoff Huston]

Hi Craig,


On 15/03/2012, at 2:20 PM, Craig Askings wrote:
> 
> Are you noticing a difference in ipv6 preference when dual stacked depending on which web browser being used? Can you even detect browsers that have just decided to stop using ipv6 even though it is available?
> 

yes, there is a difference between browsers, and a difference between OSs as well. I wrote up this in http://www.potaroo.net/ispcol/2011-12/esotropia.html late last year.

I kinda like Firefox's approach with the config setting network.http.fast-fallback-to-IPv4  set to true - it's snappy because it tries to set off both protocols in parallel.

I am still a bit wierded out by OSX Lion - it appears to keep an "average" RTT cache for IPv6 and IPv6 TCP connections and Safari will first attempt to use which protocol has the lowest running average RTT at that point (or at least that's the best explanation I can offer :-) ) The result is that even is a site is dual stack and even if for that site V6 might be faster, if the OSX client's current local running average RTT for IPv4 is lower than IPv6, then Safari will glom onto V4!

One of these days we might get it right! :-)


 Geoff