Re: Re: RR checks to avoid DoS attacks

I probably live in a simpler world than most people, but "ensure that
no packets are lost" isn't required for TCP (if you have a path
change, doing bandwidth probing isn't wrong, expecially if the
alternative has a bigger effect on measured RTTs), and isn't required
for realtime transfers (people are very tolerant of short bursts of
losses, widely separated, for both audio and video), so if we're at a
decision point, I would encourage that lossless transitions is not a
critical design goal.

Spencer

From: "Jari Arkko" <[email protected]>
To: "Tero Kivinen" <[email protected]>
Cc: <[email protected]>; <[email protected]>;
<[email protected]>
Sent: Thursday, May 06, 2004 3:59 AM
Subject: Re: [Mobike] Re: RR checks to avoid DoS attacks

>
> Good question. Looking at it technically, I agree that
> the above two cases exist. It might be another question
> whether the MOBIKE protocol needs to "optimize" for
> both cases, or whether support of the second case is
> sufficient. (There are mobility protocols that make a
> very optimized design to ensure that no packets are
> lost even if the packets have already been sent,
> but the necessary mechanisms are also extremely
> complicated.)
>
> --Jari