Minutes IETF97: v6ops
minutes-97-v6ops-00

V6OPS WG

Administravia
Note taker: Eric Vyncke
Jabber: Mikael Abrahamson
Chairs: Lee Howard, Joel Jaggli

SomeÊDesignÊChoicesÊforÊIPv6ÊNetworks,ÊP.ÊMatthews,ÊV.ÊKuarsingh
draft-ietf-v6ops-design-choices
After an hibernation period, authors have restarted work on the I-D which is
about routing-related IPv6 designs. IGP section was based on production
deployments but now it is broader. The address section was re-written to be
IPv6 only. Tim Chown: NAT is assumed to be out of scope but NPT is mentioned in
the I-D but clearly mentioned as out of scope. Getting PI addresses is always
possible but procedures differ among registries. Lorenzo Coliti: agrees with
Tim about NPT; difficult to mention ULA without talking NPT Jen Linkova: there
is I-D on how to use multi-homing with PA addresses Victor Kuarsingh: personal
opinion, PI space not so easy Tim Chown: distinction between multi-homing use
case and being independent of your ISP address space Joel Jaggli: this I-D is
for people understanding the issues of using PI space and not for general
public.

Lee Howard: let's have a couple of discussions on tweaks, then go working group
last call.

UniqueÊIPv6ÊPrefixÊPerÊHost,ÊJ.ÊBrzozowski,ÊG.ÊVanÊDeÊVelde
draft-ietf-v6ops-unique-ipv6-prefix-per-hostÊ
Comcast has started deployment this technique (even if getting less IPv6
traffic than expected). Lorenzo Coliti: using global address for NDP can cause
some issues with Apple devices, perhaps use a LLA for this traffic? This could
increase the IPv6 traffic. David Lamparter: he hopes to deploy the technique at
another conference

Lee Howard: strong consensus, so, let's go WGLC

EnterpriseÊMultihomingÊusingÊProvider-AssignedÊAddressesÊwithoutÊNetworkÊPrefixÊTranslation:ÊRequirementsÊandÊSolution,ÊF.ÊBaker,ÊC.ÊBowers,ÊJ.ÊLinkova
draft-bowbakova-rtgwg-enterprise-pa-multihomingÊ
Most changes are in section 4 (address selection algorithm). To be done: fork
another draft for the most common scenario which could be address without
source address dependent routing. Philip Matthews: different DNS could indeed
behave differently (DNS64, or CDN), so, it is a real problem Lorenzo Coliti:
one way to solve the problem is to use PvD with ID David Lampeter: about the
new I-D, why another I-D? The common use case can be solved without SADR and
could be done within V6OPS.

Lee Howard: Sounds like youÕre continuing at Routing WG, but thereÕs enough
interest here that we can keep discussing.

Local-useÊIPv4/IPv6ÊTranslationÊPrefix,ÊT.ÊAnderson
draft-anderson-v6ops-v4v6-xlat-prefix (remote)
Now using 64:ff9b:1::/48, proper IANA considerations section, ...
Lee Howard: there was a lot of discussion at last meeting: should we allocate a
prefix? Lorenzo Coliti: should provide guidances on when to use this prefix.
Tore: the prefix should be different from your own address space to make it
clear (such as not spoofing infrastructure address). David Schinazi: the new
prefix addresses my previous concern David Thaler: the section on checksum
neutrality addresses my previous concern Tariq Saraj: if we change the host ID
length, can we still do any multicast translation? Tore: no provision for
multicast currently (out of scope of the current I-D) Philip Matthews: just
scanned the I-D, why the :1 in the /48 while RFC 6052 has a 0:0 in its /96?
Tore: no worry, different addresses Erik Kline: this technique looks like
allocating ULA... Tore: but it is not subject to different rules in address
selection (by hosts)

Lee Howard: hum on adoption, which leads to adoption of this I-D by V6OPS.

Victor Kuarsingh free mike (bonus session)
Do we have interest in series of documents that describes
advantages/enhancements made possible by IPv6? Mikael A: should be very useful
because IPv6 designs do not need to repeat designs/mistakes of IPv4 Erik Kline:
spend a lot of time to repeat that IPv6 is not only 128-bit IPv4 Jen Linkova:
nice idea, Victor: could be different documents, different use cases Dan York:
really needed, and need to be specific Lorenzo Coliti: now that IPv6 is real,
it is time to use IPv6 to do more than IPv4 Tom Howard: good idea, how to
handle the issue of extension headers being dropped? Lee: could be for
specific/limited use cases Erik Kline: abondance of addresses is a key change
Dan York: I would focus on what have we learned on IPv6 Jen Linkova: some ISP
do not deploy IPv6 because they either do not know HOW or do not know WHY =>
good idea Lorenzo Coliti: we could also replace TLS SNI by using one IPv6
address per TLS server Victor Kuarsingh: we need to change how we think with
IPv6 Jen Linkova: not everyone concerned/interested is in this room (such that
operators) Bob Alderman: hopes that the I-D will be practical for enterprises

Lee Howard: Unsure whether it fits the V6OPS charter (as it is not an issue),
but, will work with the list and the AD. Joel Jaeggli: Rechartering is easy
enough. Lee Howard: We should focus on things that have actually been deployed
(even in labs). Not sure whether V6OPS is the right place to publish innovative
ideas that are untested.