Yesterday's Windows update causes IPv4 to be default

Phil Mayers p.mayers at imperial.ac.uk
Thu Nov 22 13:25:09 CET 2012


On 11/22/2012 10:54 AM, Nick Hilliard wrote:
> On 21/11/2012 12:02, Phil Mayers wrote:
>> This matches our observations. The (few) problems we've seen have been
>> services which advertise AAAAs but don't function correctly.
>
> otoh, the plural of anecdote is not data.  It would be very interesting to

Agreed. I note that our own "few" problems === small sample size === low 
significance.

I'm assuming that various browser vendors - Chrome in particular - have 
detailed telemetry that is getting collated somewhere. Presumably we'll 
see papers down the line.

> get some form of telemetry on this (via e.g. browser plugins, etc) to try
> to understand the nature + categorisation of dual-stack connectivity issues.
>
> Having said that, this issue goes well beyond browsers - although being an
> interactive medium, browsers are acutely affected by the problem.  This
> would suggest that a consistent socket API which implemented some form of
> HE behaviour might be a Real Good Idea, regardless of the underlying mechanism.

One wonders if it would be feasible or desirable to fold in some other 
enhancements into such an API - honouring SRV records, for example. But 
I agree that one simple way to do it would be a plus - and ideally a bit 
simpler than the current getaddrinfo() / connect() loop boilerplate.



More information about the ipv6-ops mailing list