Re: [Spud] States in draft-trammell-plus-statefulness-00

[Eliot Lear <[e-mail]>]

Hi Christian,

On 11/14/16 6:21 PM, Christian Huitema wrote:
> On the other hand, since endpoints may well just drop off a route
> without notice, we know that timers are needed in any case. 

Yes, needing them and waiting for them to fire off are two different
things, right?  It's better to have a mechanism that cleans up sooner
than later.  If we have that mechanism then we use fewer resources
(memory) per flow on average and that in turn means supporting more flows...


> So, we could frame the problem simply as "can we find an alternative
> to timers for state management."
>
Yes.

> My preference would be for standardizing some well-known magic packets
> with meanings like "keep me alive for another N minutes" or "drop me
> now", and to rely on bidirectionality to avoid vulnerability to
> spoofing attacks. Then, the onus would be on end to end protocol to
> incorporate or work around these magic packets.

The nice thing about TCP being stateful, however, is that the middlebox
has reason to trust how an end device is going to handle something that
is outside the state machine.  It's really well defined.  That's the
good part.  The bad part is that then the state machine is ossified.

Eliot