On Apr 18, 2014, at 10:14 PM, Erik Kline <[email protected]> wrote:
> Make no mistake though, on some access networks what they need to see
> is the client's RS, so it can get injected with Suresh's line id and
> pass up to the BNG to craft the correct customer RA in response*.
> Randomly doing DHCPv6 before even doing RS's is not guaranteed to get
> you anything in these networks.

Right.   ISTM that in some cases sending DHCPv6 messages before getting a 
response to an RS might be a minor speedup, because you can do both processes 
in parallel, but you aren't actually going to have connectivity until you get 
the RA anyway.   And if the state machines aren't interdependent, you're going 
to see all sorts of bad behavior, which I am fairly sure the Microsoft 
implementation does not exhibit.

So the fact that the Microsoft state machine has this small speed hack in it 
does not mean that the state machines are not interdependent.

_______________________________________________
homenet mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/homenet

Reply via email to