RE: [DNA] L2 link and L3 link

[Alper Yegin <alper.yegin@samsung.com>]

Hi Spencer,

> As long as we're clarifying terminology - I'm noting that we're using
> the term "L2 triggers" more frequently on this list. My recommendation
> is that we don't - we got pretty serious pushback in TRIGTRAN that
> "triggers" means something that will always be available and will
> always be acted on.
> 
> We negotiated down to "notifications", and we were headed for "hints".
> 
> I think one way to understand the difference is by saying that if you
> have a trigger, you don't need any other way to figure out that
> something happened, while if you have a hint, you would find out that
> something happened later, even if the hint was lost, so you need a
> belt to go with your suspenders.
> 
> But I'm pretty sure that calling them "l2 triggers" slowed approvals
> down by between one and two full IETF meeting cycles. Alper, Carl, did
> it look that way to you?

I think the trick is how we define triggers. If we say a trigger says
"you have changed IP subnet", you are right, this is troublesome.

Please see draft-yegin-dna-l2-hints-01.txt. We defined triggers as event
notifications from the lower layer, that say a link is brought up or
down. Pretty deterministic. Link-layer knows when an 802.11 association
is established, or a PPP link is brought up. This trigger does not say
anything regarding whether IP subnet has changed or otherwise. A trigger
may be accompanied by a hint, which might shed some light into whether
IP subnet has changed or not.

The trigger and associated hint will be consumed by the DNA mechanism.
Basically this will prompt the DNA mechanism to look into the matter.
Depending on the hint, DNA mechanism might take different actions. If
there is no hint (IEEE 802.11), a router solicitation might be sent, for
example. If the hint is the IP address configured via PPP, well there is
not much that needs to be done, maybe except a reachability test...

Alper