Re: [DNA] Re: Flash renumbering

[Erik Nordmark <erik.nordmark@sun.com>]

Sathya Narayanan wrote:

> I believe I was going for #2 - not that I clearly thought about it like
> this ;-)
> 
> I think leaving it as #1 is bad, #3 is too much and will put unnecessary
> restrictions on normal operation to address a 'low-probability' problem,
> hence we are left with #2.

FWIW I'm of the same opinion.


> Agreed. Also, the periodic RA, if it is complete will not include P1 and
> that can be used as indication also. If N CompleteRA messages are
> received without the prefix, start getting suspicious.

That is an interesting idea, because when there is intermittent 
connectivity and packet loss, it is more likely to occur for the hosts 
than for router-to-router communication (assuming the traditional 
deployment of routers on a wired network - a wireless mesh of routers 
would potentially change this).

Thus the router not hearing a prefix from another router for 1.5 hours 
is a more reasonable indication that the prefix can be dropped, than a 
host not hearing it for 1.5 hours.

But what would we do in the case when all the prefixes don't fit in one 
RA hence the complete bit can not be set?

> In conclusion, I suggest we create some text based on your suggestion
> from the previous email as a starting point.

ok

    Erik