Netflix banning HE tunnels

Sat Jun 11 00:46:36 UTC 2016

In message <m2wplwswlw.wl%randy at psg.com>, Randy Bush writes:
> > Also, the Randy who closed the ngtrans working group "declar[ing] victory" 
>  
> > yet having produced nothing.
> 
> in the ietf, that is a victory indeed! :)  from slide 9, "430 transition
> mechanisms."  the problem is they were and are a mess.  so the iesg
> decided to stop the farce.  of course, folk kept inventing new wonderful
> mechanisms, e.g. dual-stack lite, where you not only get NAT in the core
> but get to fork-lift all your cpe; a real win.

Dual-stack lite never required a forklift upgrade.  Deployment on
replacement was all that was ever needed for that.  It is a end
stage transition technology.  If you want to rush the transition
and go to IPv6 only then you need to push it out fast but a choice
you make as a ISP.

> but, underlying all this is that v6 and v4 were dead incompatible on the
> wire.  and there was/is no magic.  so all 'transition' mechanisms are by
> nature ugly, have scaling issues, sell a lot of expensive iron, ...
> 
> and it's not like folk were not screaming in pain as the ietf went down
> this insanely arrogant and stupid path.
> 
> > Dual stack is not a transition plan, and never has been.
> 
> some of us shed a lot of blood trying to explain that to deering,
> hinden, and other worshipers.  not very successfully.

But it actually worked when people let it work.

Lots of vendors delivered their piece of the strategy back in the
late 1999s and early 2000s.  Think Windows XP.  Yes, by delivering
XP with IPv6 support (yes you had to enable it) allowed every product
that ran on a Windows box to talk over IPv6 when it was upgraded.

We started delivering IPv6 capable products in the late 1990s. 

Today most equipement is IPv6 capable and the equipement will use
IPv6 if there is a IPv6 path.  How many of use actually are using
20 year old hardware with IPv4 only stacks on it.  How many of you
are using IPv4 only applications.  Software and hardware get replaced
on a pretty regular basis. I know almost all of the applications I
use support IPv6. There is very little that is IPv4 only except in
computer museums.

When you turn on IPv6 to a household +60% of the traffic is IPv6.
It would be close to a 100% if more content providers would turn
on IPv6.  It's not like the webservers they use are incapable of
delivering content over IPv6.

What you should be doing is complaining to the people still selling
obsolete IPv4 only garbage.

Could it have been done better, sure.  We could have lobbied for
governments to ban the selling of new IPv4 only equipement (except
with special dispensation).  That strategy worked for the imperial
to metric transition in Australia.  You could not buy new imperial
only equipement down to rulers at the local newsagencies.

> > It's a key factor in why we have such a fractured adoption today.
> 
> this i do not buy.  dual stack allowed some backbones to get v6 to the
> edge.  this did not fragment adoption, it was just far from a scalable
> total solution.  then again, nothing else is very pretty either.  the
> tragedy is that there are now more folk using cgns than ipv6.  the nat
> haters have created the worst nats in the world; justice, but not the
> kind of justice i like.
> 
> > If you've been completely ignoring IPv6 for 20 years
> 
> have not, though i suspect our bean counters wish we had.  it's been a
> very expensive road.  our first deployment was on a truly dual stack
> backbone, separate circuits and separate routers (netbsd boxes as v6
> traffic was small) as j&c did not support dual stack, heck any stack.
> 
> randy
-- 
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742                 INTERNET: marka at isc.org