Re: [dhcwg] Please review version -07 of draft-ietf-dhc-dhcpv6-stateful-issues

["Bernie Volz (volz)" <[e-mail]>]

Jinmei:

We have to handle EXISTING servers which I believe base the T1/T2 times on each IA separately - though that is just my assumption -- I haven't analyzed all of the servers. Obviously this also depends on the lifetimes configured (i.e., if they are all the same and the server assigns the addresses/PDs are the same time, the lifetimes should generally all be identical and hence so should the T1/T2 times).

This is the point about the client text - it is there to handle the case where the client is communicating with an older (3315 compliant) server. If we required all servers to be magically updated, we wouldn't need the client text.

And, yes, I was not trying to be too explicit about this because how a client choses to implement this is up to it. For example, it might just schedule timers to expire for each IA based on its T1/T2 times. But when the first timer goes off, it includes all bindings from that server in the Renew. So, in this case the client has no need to "compare" the T1/T2 times at all.

Perhaps that is a way to solve this problem - perhaps the text should focus on the behavior to say that:

"The client SHOULD Renew (or Rebind) all of the bindings at once, whatever the individual IA T1/T2 times might have been. How a client handles this (whether it determines one set of T1/T2 values to use across all bindings, initiates independent timers and the first one to go off results in a Renew (or Rebind) with all bindings, or some other approach), is up to the client implementation."

- Bernie

-----Original Message-----
From: [e-mail] [mailto:[e-mail]] On Behalf Of ????
Sent: Friday, October 17, 2014 1:33 PM
To: Bernie Volz (volz)
Cc: [e-mail]
Subject: Re: [dhcwg] Please review version -07 of draft-ietf-dhc-dhcpv6-stateful-issues

At Thu, 16 Oct 2014 19:08:07 +0000,
"Bernie Volz (volz)" <[e-mail]> wrote:

> If the client sent a Renew and the Reply has 0 lifetimes for EITHER (but not both) the IA_NA or IA_PD. What should the client do?

I've not considered all of the ramifications (and, frankly, I've not read all of this response closely), this breaks basic operational assumption (i.e., all lifetimes and T1/T2 should be same for all IAs), right?  If so, we might rather not try to be perfect but just make sure that client ends up with a valid state (like having T2 < T1) and leave other details open.

--
JINMEI, Tatuya