I don't think the issue is so much the lifetime of the RA as the need to
wake up a dormant mode host when an unsolcited RA is multicast. I assume a
dormant mode host coming out of dormant mode will have some amount of work
to do to get its IP traffic channel in shape. This will naturally include
using DNA to solicit an RA, since the host may have moved to a different
subnet while it was in dormant mode. Am I missing something?
jak
----- Original Message -----
From: "Syam Madanapalli" <[EMAIL PROTECTED]>
To: "Erik Nordmark" <[EMAIL PROTECTED]>
Cc: <ipv6@ietf.org>
Sent: Thursday, August 10, 2006 12:31 PM
Subject: Re: Proposal to change aspects of Neighbor Discovery
Hello Erik,
On 8/11/06, Erik Nordmark <[EMAIL PROTECTED]> wrote:
Syam Madanapalli wrote:
> I am not sure if this works.
> Let us say the router lifetime is X seconds
> and the host wakes up every Y seconds to retrieve
> any packets from the AP/BS.
> If Y > X then we are not solving the problem.
I think you need to explain what would fail in that case.
Sure, DNA helps, I did not thought about it earlier.
But it would be better if the lifetime of the default router
is greater than the sleep time, which could avoid the
DNA operation.
In my simple view of the word, when the host wakes up the software is
told (or notices by watching the clock tick) that it has been sleep for
4 hours.
It then notices that the router lifetime expired one hour ago.
Instead of just removing the router from the default router list and
doing nothing else it invokes DNA (which in its simplest form is to send
a RS and wait for a RA). FRD can help shorten the time to get the RA
assuming the AP/BS knows from L2 when the host wakes up.
Thus the host would just apply a bit of slack time after it has woken up
before it starts timing out state (that would normally have timed out
while the host was sleeping.)
This doesn't change any protocol; for IP on the access router it looks
like the host disconnected and reconnected.
> Even if we want use FRD, it still useful to increase
> the router lifetime hence the periodic RA's interval.
Sure. Increasing the periodic RA interval wouldn't hurt, especially if
the host can detect when it might be connecting to a different AR (using
DNA).
Yes.
Thanks,
Syam
Erik
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
