Hi Erik,
Erik Nordmark wrote:
If Optimistic DAD doesn't allow for unicast responses to router solicitations, the potential for fast advertisement to such hosts is severely diminished.
So how could something work?
I'm assuming that somehow, perhaps using a token bucket filter instead of a fix 1 every 3 seconds limit, we can accomplish faster RA responses to RSes, but the DAD interactions are still perplexing.
Alternative 1: The RS contains the tentative address as the source, but no source link-layer address option (since that would override any link-layer address recorded in the routers in the case of an address conflict).
Thus in order for the router to unicast the RA it needs to perform a NS/NA exchange with the host. This is only done when the router doesn't have a neighbor cache entry, and the resulting NA has override=0. So this will never replace a cached link-layer address in the case of a conflict. But the exchange is 4 packets instead of 2.
Indeed.
This is what is already working in router implementations when the SLLAO is not present.
The 4 packet exchange is not ideal. I'd prefer to get the exchange down to 2 (RS/RA).
Alternative 2: Why not conceptually define the override flag for the RS to accomplish a 2 packet exchange? A compatible way to do this is to define a new "tentative source link-layer address option" which would only be used by the receiver if it doesn't have a cached address i.e., would never override an existing address. (As an aside, this could also be used for NS packets.
I've seen your proposal before (was it on a mailing list?, or in an e-mail round with JinHyeock?).
It's a good idea since the fallback position (where the option isn't interpreted by the router) works in the same fashion as Alternative 1.
I'll contact you on aside to see if we can cover the salient issues and describe how such an idea could work reliably.
Greg
-------------------------------------------------------------------- IETF IPv6 working group mailing list [EMAIL PROTECTED] Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
