> My customers need implementation of 3.3.1.4 and asked for it.
The argument is steel :-)
Well, the desire of customer is law, but believe me.
I gathered statistics of router failures for two years
and found that time when routers stop to answer arps
does not exceed 2% of real gateway failures, when they
stop to forward. It is the case both for Cisco routers
(mainly, maintanance time) and Linux routers (mainly,
gated crashes).
Sell your customers rdiscd, they will be happy :-)
> I think the proposed algorithm will work and is easy to implement,
> so I will give it a try.
It is easy. Really, 2.1 kernels already have it in latent state.
Alexey Kuznetsov