Minutes for SUNSET4 at IETF-95
minutes-95-sunset4-2

Sunset4 session, IETF95
5 April 2016, 17:40

Scribe- Fred Baker (augmented with subsequent notes via Lee Howard reviewing
session recording) Jabber- Mikael Abramsson (Jabber logs:
https://www.ietf.org/jabber/logs/sunset4/2016-04-05.html)

- administrativia

- IPv4 Declared Historic, Lee Howard draft-howard-sunset4-v4historic-00

Lee essentially read the draft to the working group, all 500 words of it.

The author argues that this is true of IPv4 as a matter of course; we are just
commenting on the fact.

Issues raised:
Fred Baker: "Deprecated": see RFC 1158 section 3.1 for a definition. We need to
decide whether it applies, but there is in fact a definition.

Dave Thaler: “Historic” doesn’t mean “can’t update;” it says that “standards
track documents <i>normally</i> must not depend on (documents) at a lower
maturity level.” He emphasized “normally” and gave examples of exceptions, and
concluded that declaring IPv4 Historic would not prevent updates to IPv4, it
would just raise the level of scrutiny by the IESG. He therefore agreed with
declaring IPv4 Historic.ar

Fred Baker: Many RFCs use the term "IP" to mean IPv4. Deciding that it should
now mean IPv6 forces a new interpretation on many RFCs that would do violence
to the clearly intended meaning.

Philip Hallam-Baker comments that deprecating IPv4 will be perceived as trying
to turn it off, and invites ridicule. He strongly opposes doing something that
invites ridicule. He warned against abandoning development on IPv4 to another
SDO. He further said that he had no interest in anything that only works on
IPv6.

Ruediger Volk: agrees and thinks the declaration is counter-reality and not
strategic.

Geoff Huston: the purpose of an SDO is to create technologies for business
purposes. Not to pick winners or have opinions, but to write specifications.
Example of previous tech where IETF decided not to do work was NAT. We ended up
with lots of people being "creative" and are in much worse shape as a result.
Another SDO might step in if IETF decides it's done with IPv4. “We’re
fundamentally changing the semantics and behavior of v4,” Geoff continued, “A
v4 address is not a permanent endpoint identification token any more.  We’re
now looking at temporal conversations that support conversations, and if I
understand QUIC correctly, we’re even over that: all I want is for the v4
address to remain permanent for one RTT. . .  Necessarily, I think, as we
change the semantics of the IPv4 address, we might need to think about
specifications that others can rely on that we can interoperate with that will
help us to the impossible, and cram that many in.”

Wes George: Not all protocols require a given version of IP. The question is
how long we need to retain 100% backward compatibility. The IETF’s timescale is
long, because we define the standards implemented and deployed on the Internet;
whenever we do this, it will look premature to the rest of the world, because
it will be a forward-looking statement whenever we do it.

Barbara Stark: Doesn't think this makes sense for her business.

Aaron Falk: "We can't do this because we have an enormous user base".

Phillip Hallam-Baker: an example of a technology this would negatively affect
is TCP Crypto. A better status might be "it's complicated".

Nalini Elkins: We can talk about what we might do. It was suggested verbally
that we might tell other SDOs to stop using IPv4 (which the draft doesn't
actually say), and we don't have authority to say that.

Eric Nygren: Need to use a different term that basically means "end of life".
This is about the optics of the statement outside the IETF. The time is at some
point that IPv6-only deployment is becoming common, and IPv4-as-a-service is
worth considering.

Comments from Jabber: do we need a status from the IESG that would mean "we
don't want to do new work" on IPv4? We might want to make a statement to other
SDOs and partners supporting IPv6.

Ruediger Volk: observes that we aren't using IPv6 in examples in IETF, default
networks aren't IPv6-only.

Brian Haberman: possible for community to push on WG charters to ensure that
IPv4-specific work doesn't make it into charter

Petr Lapukhav: People still see IPv4 as a long-term thing. Facebook would like
to see IPv4 development discouraged, as they have stepped in the IPv6-only
direction.

Ted Lemon: An endeavor like this needs to be community-driven, and widely
agreed to.

Marc Blanchet: Issues with other specifications, such as IPv4 literals in
URL/URI. Very concerned about dependencies on IPv4 such as html, SIP.

Mark Townsley: We need to focus more on making IPv6-only viable and set it up
as "Internet" service. expressed concern about fragmenting the Internet, if a
brave operator turns off IPv4 and their government says that’s anti-competitive
against IPv4-only content. Lee Howard: described the reverse, where a regulator
thought that the lack of availability of IPv4 was considered a barrier to
competition for new market entrants. Mark Townsley: proposed that a statement
describing the need to transition to IPv6 only would provide support for such
operators.

Will Ivancik: Worried about what this really means.

Barbara Stark: Agrees with what the draft is trying to do, but observes that if
something needs to get something done will find a forum or avenue to accomplish
it.

David Schinazi from Apple: The only way we will deploy IPv6-only requires
transition technologies. These may not be perfect, and may need fixing.

Mikael Abrahamsson: there is great value in a statement of direction towards
deprecating IPv4.

Tom Herbert from Facebook: we need a symbolic statement, and "historic" isn't
it. Need a plan for success in IPv6 deployment. Describing the end-of-life
path, and metrics triggering each step, would be useful.

David Schinazi: It would be nice to have a transition technology that works
well. We don't. Need a better statement of intent.

Ruediger Volk: Avoid technical fragmentation of the internet. Call it "End of
Engineering" (as opposed to "End of Life")

Phillip Hallam-Baker: minimum time to being able to change status of IPv4 is
five years.

Igor Gashinsky: "There's no way we're actually going to make IPv4 historic
right now". We could say that we would like to take it to end of life in a
decade, it't not something we will actually do. We cared about 0.02% of our
users breaking before we could enable IPv6, why should IPv4 be different in
terms of when breaking it stops mattering?

Mark Townsley: David, when Apple decided to push IPv6-only, was that based on
something the IETF said or did?

David Schinazi: We did it because it made sense in our environment.

- next steps
Lee to update draft based on feedback
more discussion on list about what the right direction is, including IESG and
IAB statements about drafts, Liaison statements to other SDOs, etc.