intarea WG
IETF 89
TUESDAY, March 4, 2014
1300-1400 Afternoon Session I
Minute takers: Charlie Perkins and Emmanuel Baccelli
========================================================================

- RFC 7098 published (flow label balancing)
- Brian: closed the IP and DNS directorate, started a new directorate to
    assist the ADs with reviews
    = in the process of populate directorate
    = send nominations
- GRE Fragmentation
  = Goals of draft: does NOT update RFC 2784.
  = Specify a GRE tunnel gragmentation strategy
  = Strategy 1: discard packet (does work)
  = Strategy 2: Fragment payload (only works for IPv4)
  = Strategy 3: Fragment delivery
  = Current behavior: fragment payload if possible, otherwise discard & ICMP
  = Applicability: over IPv{6,4}, PMTU *must* be IPv{6,4}_MTU + GRE_Overhead
  = Mark Townsend: GRE header is variable, so what does GRE overhead mean?
    == In draft, means IP header plus *longest* GRE header in RFC 2784
  = Bob Hinden: Recommendation is same for IPv4 and IPv6
    == If delivery header is IPv4, IPv4 MTU is much smaller
  = Dave Thaler: how much of the draft is GRE specific, versus applicable
    to all tunneling protocols?
    == Answer: absolutely right
  = Mark Townsend: actually have document from Joe Touch which is actually
    WG document; should pick this up again
  = Mark Townsend: there are THREE documents specifying GRE (incl. 1701)
  = Suresh: previous document was abandoned.
  = Mark Townsend: this is a good way to motivate picking up the document
  = Ron: this draft puts out of scope the cases where fragmentation is
    already required.

Baccelli draft:

Data Plane Acceleration / Zao:
- CAGR of Global Internet is 24%
- Virtualization technology increases traffic within appliances, magnifying
  the "forwarding gap".
- DPDK / accelerated virtual switch / www.opendataplane.org
- Fast Path within one device
- The higher layer of processing required for a packet (e.g., L4-7),
    the worse performance
- Acceleration relationship to IETF standards:
    multiple devices + communication between control & data plane:
        wire-visible protocol is indicated...

========================================================================
draft-baccelli-manet-multihop-communication (notes by Emmanuel)

Erik Nordmark: would be interesting to document some of the solutions
that were developed in various was to deal with that issues.

Thomas Clausen: there is a difference between routers and hosts.
We should consider this more precisely in this draft. There were
long discussions about this distinction in AUTOCONF that led to nowhere.

Emmanuel Baccelli: This document is agnostic to this distinction because
it focuses only on basic characteristics of wireless communications.
The aim is to provide a reference should anyone want to design Internet
protocols over such networks. The goal is precisely to help avoid
future fruitless discussions in other working groups.

Robert Cragie: I disagree with Thomas. The characteristics described
in the draft are agnostic to the router/host distinction.

========================================================================


Glen Wiley / Verisign / getdns-0.1.0 release
- Paul Hoffman edited an app-oriented DNS API, April 2013
  = access to DNSSEC, fully asynchronous, eased access to DANE, SRV, ...
- OPen Source Implementation from Verisign + NLnet Labs
  = including QA engineers
  = BSD-New license, you can do whatever you want with it
  = Creative-commons like
  = opendnsapi.net
  = libldns and libunbound from NLnet labs (libldns requires openssl)
- Major features of this release (it compiles!)
  = wiki details parts of Paul's specification that are not implemented

========================================================================

Erik Nordmark / CGA-TSIG
- Secure PTR update
- Draft is not implemented on OpenDNSSEC

========================================================================