SAVI, Nov. 9, Tuesday, Beijing

Minutes taken by Guang Yao, yaoguang.china@gmail.com



1. vogt: presentation sequence changes

Three main issues:
(1) data triggering and binding removed; term change
(2) add people involving from very beginning
(3) deliverable

open question: include mixed DHCP/SLAAC case also in framework?
a new separate doc:
default: not include

Marcelo: what exactly do you think? what would that be?
V: a section of very high level abstraction, e.g., priorty configuration.
Marcelo: too much details will delay the deliver of the framework.
V: disscusion will be limited.
M: will manually configuration be included?
V: something in general will be discussed but not specified solution.

(1) Jun Bi, cernet deployment of savi
implementation: dhcp-06, stateless-01, mix-00
vendor: ZTE, huawei, H3C...
console example
100 univ.
student buildings, 3 scenarios
sample universities
MIB and management system
SAVI and user anthentication

Marcelo: in mix scenario, how do you solve collision between DHCP and SLAAC?
Jun: in mix-mode presentation

(2) Jun Bi, solution for DHCP
major revision: modified to DHCP-only scenario

V: it has been a mature document and it is suitable to adopt it for wg last call

(3) Marcelo, SAVI-SEND, for SEND only
no comment

(4) Marcelo, SAVI-SLAAC

Bi: idle host problem; busy host will lose ownership of address if not replying to DAD NS.
M:  which doesn't response? mouse? generally it's high level situation.
Bi: I mean common situation.
Joel: not serioues problem. The problem is from slaac, not from SAVI. Use DHCP or manually configuration
B: no SAVI and SAVI make different. SAVI makes thing worse.
J: SAVI is not only the case to fail. a number of mechanisms will fail because of unpredictable error.
M: it is not for IPv4(?). It costs the attacker a lot to succeed in making host idle. It is not a real success.
B: attacker can try many times, repeatly.
V: What would a solution to that problem look like? Current solution is for mobility support.
B: you mean how to handle this case? yes. A delay timer to hold binding.
V: It will defeat mobility.
J: If using iphone and moving very fast, timer doesn't work.
B: just a short timer.
M: if using a longer timer, savi behavior will not be the same as host behavior.
Erik: if host sleeps, it is hard to predict the time. there must be an identifier to distinguish if same host moving. 
			It is better than no savi at all. Strong security is of cost and out of the scope.
V: different secrutiy property in different solution. A better solution is another address assignment method. Not a problem of SAVI, but a problem of assignment method.
B: anyway it cannot be ignored.
B: one more: IPv4 should also be considered.
M: this doc is for ipv6 only. ipv4 later, a new doc.
V: this doc is for ipv6 slaac.

(5) Jun Bi, Mix

M: not a network. The same link.
Joel: I manage a slaac network. DHCP server is also deployed. Some users update the OS to support DHCP. 
	 However, I can not wait all hosts update to DHCP. This is a user case for mixed scenario.
Erik: it still works well if same address bound by different solutions.
B: we don't want same address bound on different anchors.
Erik: same address on different anchor? it is not a savi problem but address assignment problem. 
	    we don't want savi to make things work, but to improve security.
M: this is the order?
B: next page, just a list
M: slaac and CGA are the same?
M: CGA is much stronger than slaac
B: this is an initial edition. discussion can be continued
M: how to define same prioty? if same, look at DAD only.
Jari: DHCP address will be declined when DAD fail
B: DAD check is removed from DHCP doc.
Jari: real example? I don't know the inner of solution
V: at the end of DHCP, verify DAD anyway?
B: if same priorty is set for DHCP and slaac
Jari: FCFS is ok.
M: best is look at DAD
Jari: the red part. do you follow, or provide priority?
B: follow, except DHCP and slacc.
Jari: do need something more? follow DAD is enough.
Eric: change priority may bring problem to host...
Erik: before change ...(?not clear)
Eric: if blocking a previous binding, let the previous guy know. The best is making constrain on the DHCP space.
V: There do have DHCP and SLAAC mix use cases. DHCP also performs binding procedure slaac, to follow slaac seems enough. 
	address this question in the new revision.

Bi: adopt to i-d?
V: i want to see a new revision. questions from wg should be resolved 

(6) Yilan ding, PANA for SAVI
No comments.

(7) Tao Lin. NDP Roaming
V: the perimeter concept already makes the savi scalable.

(8) Joel. threat scope