Re: [DNA-BOF] Comments on draft-kumar-dna-reqmnt-ipv4-00.txt

[Alper Yegin <alper@docomolabs-usa.com>]

On 10/24/03 1:06 PM, "Bernard Aboba" <aboba@internaut.com> wrote:

>> Couldn't a DHCP server act upon a "link-layer trigger" and release the
>> resources? Unless this is prohibited in the DHCP specification, I suspect
>> this can very well happen.
> 
> I'd suggest that attempting to build a DHCP Server that kept track of
> "link down" events everywhere on the network and revoked resources as a
> result would be a very bad idea.
> 
> The result would be a DHCP server that was constantly revoking leases
> without client knowledge.  "Link down" events are very common,
> and most of the time are quite harmless so the best thing to do is to
> ignore them most of the time, not "act on them immediately" as was
> suggested in this requirements document.  Since clients routinely ignore
> these events (to their credit), if the server were to attempt to take them
> into account the client and server would be out of sync much of the time.
> 
> Overall, there is a very good reason why the idea of remoting "link down"
> triggers has not progressed.

Checking the RFC2131, I see the "deliberate design" that practically
disallows DHCP server from revoking the lease without the client's
knowledge.

On the other hand, imo the weight a "link down" event carries really depends
on the architecture/access technology. For example a tear down of a PPP link
in 3GPP2, or PDP context in 3GPP, or an ATM circuit in DSL is not equivalent
to a disassociation in IEEE 802.11. The former is less likely to
(unintentionally) happen during a session, and more closely tied to
configuration management.

Alper