Re: [savi] 答复: Binding establishment based on data or control packets?

[marcelo bagnulo braun <marcelo at it.uc3m.es>]

yaoguang escribió:
> I mean, SAVI bindings are soft state, hence it can and will be discarded 
> at some point. 
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> I don't quite understand your meaning of " SAVI bindings are soft state ".
>   

What i mean by this is that the state is created by the SAVI device as
the response of a data or ND packet sent by the host (but that the goal
of that packet is not to explicitly create state) and that the
responsability of discarding that state is completelly up to the SAVI
device i.e. the host will not send a FIN packet to delete the state.

In other words, the SAVI device will receive a packet from the host, and
will create the SAVI binding
The SAVI binding will have a given lifetime.
It is likely that the lifetime is extended if the host keeps on sending
packets, but if a long period passes and no packets are seen by the SAVI
device referring to that address, the SAVI device will unilaterally
(i.e. without asking the host) will discard the state. This is
essentially needed for memory reasons

I guess a reasonable lifetime for a binding would be similar to the dhcp
lease lifetime... so, between a day and a week, would seem reasonable to me

(probably, if the SAVI device is under a DoS attack, it will need to
discard the bindings earlier)

So, it is perfectly possible that a host is the rightful owner of an
address and that the SAVI device has discarded the binding

regards, marcelo

> When and why will SAVI bindings be discarded, and how often?
>
> I think they are somewhat soft, but they are not as soft as bubbles.
>
>
>
>