Tina, Thank you for your comments.
Hi Erik and Igor, The draft is well explained in regards to the case when there is no alternative default router in the lest of default routers. I wanted some clarification in the content below: In the draft, The recommended behavior is to have 5 attempts, with timing spacing of 0 (initial request), 1 second later, 3 seconds later, then 9, and then 27, and switch to UNREACHABLE after 3 transmissions, which represents:-------after the initial request, we transmit a packet after 1 sec, then another after 3 seconds, then another after 9 seconds and finally after 27 secs right?
The times in the example are relatively to the previous (re)transmission. I've expanded the example to also show the cumulative time since the first transmission. As part of doing that I also formalized the cap on the exponential backoff (suggested in the text as 60 seconds) by introducing a MAX_RETRANS_TIMER.
Your questions below indicate to me that the text is confused since in some places we say that we switch to multicast and mark UNREACHABLE at the same time. But the example was intended to show a case when those occur at different times.
MAX_UNICAST_SOLICIT=5----This would imply 5 attempts?
Yes, 5 unicasts.
RETRANS_TIMER=1 (default) BACKOFF_MULTIPLE=3 MARK_UNREACHABLE=3------Do we mark it unreachable after 9 secs OR 4 secs 0r 27 secs?
After 1 transmission plus 2 retransmissions. With the recommended timer and backoff values that occurs at 4 seconds since the initial transmission.
After marking unreachable, multicast NS will be send instead of Unicast NS right?
The (non-normative) example is slightly more sophisticated in that it first marks as unreachable and later switches to multicast. That allows for more quickly switching to an alternate (e.g., in a case with multiple default routers), while reducing the impact of multicast packets. The downside of this is that it takes longer to detect a node (host or router) which keeps its IPv6 address while changing its MAC address.
Since there will be 5 attempts, would it imply it will wait 27 seconds before declaring it unreachable?
No, in the example (which was confusing and contradictory) it was supposed to be after 4 seconds, while discarding the NCE (implicitly switching to multicast NS) after 40 seconds after the initial transmission.
I think I finally got this right in version 05 that I just posted. I'd appreciate if you can look it over for clarity on these matters. Thanks, Erik -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
