Brian,
This is hardly to prove a point.
It is to have an operational experience so that stuff gets fixed and standards
get written to fix stuff.
An IPv4 outage is radical, granted, may be it should be done during hackday on
the weekend before IETF. May be it should not be done at all and instead the
default network SSID:IETF to be NAT64, with an SSID:IETF-IPv4-only.
May be NAT64 is not the right one, may be it is XLAT464? (NAT64/DNS64 does not
require much client participation, XLAT464 does).
May be it is way too short to have it in Chicago? Likely.
May be it should be better planned. Certainly.
We need a transition to IPv6-only, this go via dual stack, sure, but do we have
to wait for the whole world to be IPv6 before we can deprecate IPv4? No, we
need something that allows to have IPv6-only networks (enterprise, guest,
customers,...) capable of talking to the legacy Internet. We also realize that
the bottom line is: "dual stack=dual costs" and "dual stack=dual security
issues". As the world progress, the need for NAT64 will go away...
So, this is a very interesting discussion, as long we move towards some
solution(s).
What do you propose?
----- Original Message -----
From: "Brian E Carpenter" <brian(_dot_)e(_dot_)carpenter(_at_)gmail(_dot_)com>
To: "Franck Martin" <franck(_at_)peachymango(_dot_)org>
Cc: "IETF" <ietf(_at_)ietf(_dot_)org>
Sent: Tuesday, January 24, 2017 5:21:23 PM
Subject: Re: IPv4 outage at next IETF in Chicago
Franck, I try not to be religious about NAT, and I use NAT44 every day
like most people. Also like most people, I experience occasional
unexplained failures of web-based transactions. Whether they are due
to a NAT garbage-collect or a load-balancer failure, I don't know,
of course. But actually I am not deeply concerned about NAT64, although
any failures that it generates would be very hard to identify. I am
more concerned about IETF participants whose devices are not set up
as dual stack nodes at all. They would be completely blocked. Yes,
I know, such people should not exist, should be deeply ashamed, etc.
But I don't see why we would cut them off to prove a point.
Regards
Brian
On 25/01/2017 14:05, Franck Martin wrote:
----- Original Message -----
From: "Brian E Carpenter"
<brian(_dot_)e(_dot_)carpenter(_at_)gmail(_dot_)com>
To: "Franck Martin" <franck(_at_)peachymango(_dot_)org>, "IETF"
<ietf(_at_)ietf(_dot_)org>
Sent: Tuesday, January 24, 2017 4:33:22 PM
Subject: Re: IPv4 outage at next IETF in Chicago
On 25/01/2017 12:11, Franck Martin wrote:
I think it is time to move to the next level of IPv6 deployment.
Ideally the IETF WiFi network should now only provide the following 2
networks:
1)IPv6-only
2)IPv6-only with NAT64
The later should be the default network.
However you would say, well some stuff will break, some non technical people
will use the IETF network and may have a bad experience, etc...
So to be conservative but at the same time futurist and like it was done a
few
years back, why not create again an IPv4 outage of a few hours where the
above
2 networks would be the only networks available?
That would be a good way of damaging IETF productivity for a few hours.
Do you have evidence of applications not running in a NAT64 environment? I'm
interested to know them.
And for what? Moving away from the mainstream coexistence mechanism (dual
stack),
to a mechanism known to be intrinsically defective (NAT). I don't see the
point.
I fail to see how NAT is intrinsically defective, since it is used
successfully by everyone...
Nevertheless, the goal here is to get the Internet designers (IETF) to have
operational experience on what needs to be fixed.
When the IPv4 outage happened a few years back, it gave a serious impetus in
getting IPv6 totally mainstream on many platforms.
IAB encourages IPv6: https://www.iab.org/2016/11/07/iab-statement-on-ipv6/
However going IPv6-only can only be done in walled gardens. There still will
be many environments with IPv4 only. A solution here is to move networks to
NAT64, so you only need to support IPv4 at the edges...
Yes creating an outage for the sake of an outage is pointless, experience on
what works and not work needs to be recorded.
May be the first step instead of doing an outage is to have as default a
NAT64 network at IETF meetings and a dual stack network for the people that
experience issues.