Introduction and Document Status, Chairs, 10 min

no comments on agenda/status

Working Group Drafts

Tommaso Pecorella: what if information keeps being sent and lost?
maximum time limit before giving up?
Fernando Gont: this is assumed to be configurable, could default to
3, and at some point SLAAC just breaks
Tim Winters: Multicast RA before getting rid?
Fernando Gont: no, because per router basis
Jen Linkova: send All-Routers RS instead of unicast? e.g. router
changes address
Fernando Gont: expect that router multicasts RA after address change
Jen Linkova: removal - on invalid or deprecated?
Fernando Gont: would just be gone when the last router disappears
Jen Linkova: breaks connectivity for remaining hosts?
Fernando Gont: (missed on notes)
Jen Linkova: does "no response" include "response not including PIO"
Fernando Gont: would include the latter
Bob Hinden: never seen router splitting information, is it useful?
Fernando Gont: not seen, algorithm is just robust against it
Erik Kline: TBD section 4.5 - state machine or narrative? preference
for state machine
Fernando Gont: mostly ready to send out to look at

Éric Vyncke: repeat of previous comment - fixed length option for
single use case is a waste, would prefer varying length for multiple
use cases
Jie Dong: The rule in hbh-processing draft suggests using fixed
length options, which may need to be considered. We've introduced
new fields to improve extensibility
Joel Halpern: intention/wording is confusing, hoping for resource ID
to identify allocation, but not clear. and relation to teas NRP ID
header should be clarified.
Jie Dong: yes, intended to be resource ID. While further discussion
on semantics generalization may have impact to this.

(chat comment missed)
Éric Vyncke: annoyed by using echo request/reply for this, would
prefer e.g. Node query
Xiao Min: (? / not understood)
continuing on list

Darren Dukes: SR source knows destination, compressed does not
matter, and by the time packet reaches dest, address will be
replaced - no need for this draft, works as is
Xiao Min: have problem statement in draft
Darren Dukes: disagree with problem statement
continuing on list

David Lamparter: anything achievable by changing selection order can
also be achieved with additional LL address as nexthops
Paolo Volpato: open to discussion
Erik Kline: ACKing 8801 per-PVD link local addresses solve this,
it's done, issues with that?
Fernando Gont: overlap with other WG & v6ops documents, need to do
this in that context; document also tries to address operational
concerns, that should go to v6ops; how about noting down the
remaining problems/gaps first?; also lots of very specific cases in
current doc rather than holistic approach
Eduard Vasilenko: tried to delete everything that overlaps, trying
to solve specific case, will take feedback into consideration
Eduard Vasilenko: believe 8801 to not be enough, planning to discuss
offline
Jen Linkova: seconding PVD comment, topic started in 2016, if PVD
does not solve this that needs to be clarified
Eduard Vasilenko: PVD only good enough after source address was
chosen, not enough on its own. but yes PVD is applicable.
Jen Linkova: section in beginning of document to explain this

comments on slide 4

Éric Vyncke: again binding IID to MAC - back to EUI-64 problem and
lack of privacy?
Eduard Vasilenko: no, enough information fed into hash; will chose
different inputs
Éric Vyncke: madinas WG about randomizing and changing MAC address,
being deployed, how does it affect this?
Eduard Vasilenko: generating new identifier should still work with
this
Éric Vyncke: should be mentioned in draft

comments at end

Fernando Gont: goal is to prevent IP address impersonation?
Eduard Vasilenko: yes, nobody else can claim address
Fernando Gont: what are the attacks you have in mind?
Eduard Vasilenko: slide 2, poisoning caches of other nodes, MITM
Fernando Gont: tying identifiers has generally proven to bring
problems, move away from EUI-64 similar, and if you need RA-guard
anyway why not ND inspection to fix this?
Eduard Vasilenko: not enough support for ND inspection in switches,
RA guard simpler & supported everywhere. CGA/SeND connecting public
key to IP, this is connecting MAC to IP instead. Avoid need for cert
infrastructure.
Jen Linkova: not clear why only disconnected node could be attacked
Eduard Vasilenko: partially true, assuming L2 infra would flap when
same L2 address is reused
Jen Linkova: relying on monitoring for this, not prevented?
Eduard Vasilenko: yes, assuming flapping, noticeable issues
Jen Linkova: address conflicts could be detected too. / Separately,
if hosts are not trusted, why not deploy L2 isolation?
Eduard Vasilenko: L2 isolation not typical in targeted situations,
generally rare
continuing on list

Peter Psenak: when flexalgo was designed, requirement was not to
have any specific date in the packet, working on label or address.
what's presented here is completely opposite to that. tried and
failed due to classification at every hop.
Éric Vyncke: same comment as on previous draft, VPN ID marking is
interesting, but make it generic
Suresh Krishnan: is the host adding this information? who is putting
it in?
Zhibo Hu: ingress router doing the encap adds it
Suresh Krishnan: security & related properties are scary, needs more
consideration & discussion
Zhenbin Li: design of flex algo consumes more IP addresses, trying
to avoid that with alternate signaling. also active discussion in
teas WG about identifiers
Joel Halpern: does not add up; needing IP addresses is a benefit,
not drawback, enough addrs in IPv6. new extension header to take
into account is a very complicated solution to a non-existing
problem. please don't.
Tommaso Pecorella: missed something but extension header cannot be
added by router?
Zhenbin Li: router is adding encapsulation header, can add extension
headers as part of that
Jen Linkova: ACK with Éric, lots of different drafts for this
slicing/identifying type of thing, should be merged?
Erik Kline: mailed teas AD re. status of network slicing documents

draft presented but insufficient time for comments

no time left to present this draft

no time left to present this draft

no time left to present this draft