Re: [Anima-signaling] GRASP issue 50: Rapid mode

[Toerless Eckert <[e-mail]>]

If i understand the message names correctly, then i think what we
discussed in Berlin was as follows:

- The registrar objective needs to be discovered by every autonomic node in the
   network (because every autonomic node can be bootstrap proxy).
- It would be too much signaling traffic if every autonomic node was sending
  discovery messages. Trying to optimize this would also depend on topology,
  so could be difficult.
- The easiest solution is to define the option for "densely used objectives"
  (like registrars), to periodically send out unsolicited "Discovery Response"
  messages. These "Discovery Response" messages would have to
  - Include Rapid Mode, aka include the Synchronization information
  - Use hopcount MAXHOPCOUNT so they are flooded throughout the network.

Aka: Rapid mode in that case must not be limited to link-local.

In general, i would keep complexity of first RFC as low as necessasry for the
use-case examples we feel we need to support:

- Whenever GRASP does unicast-negotiate between objetive and client, rapid mode
  seems to only provide reduction by 1 round trip. I know how hard web-browser
  vendors like to work to do this with eg: QUICC, but it took them 2 decades to
  arrive at that performance level requirement. I don't think we should introduce
  all this complexity right now unless we have clear near-term requirements.

- Aka: I could happily let go of rapid-mode for now in the general case
  (move into appendix as "candidate extension"), and just figure out how
  we can include unsolicited synchronization data into unsolicited discovery
  messages.

Cheers
    Toerless

On Mon, Aug 01, 2016 at 08:58:12AM +1200, Brian E Carpenter wrote:
> Hi,
> 
> Here's another issue for design team comments.
> 
> My understanding from the discussion in Berlin is that we should retain the Rapid mode
> but also clarify its specification a bit. At the moment the open issue says:
> 
>    o  50.  Is Rapid mode limited to on-link only?  What happens if first
>       discovery responder does not support Rapid Mode?  (Section 3.3.4,
>       Section 3.3.5)
> 
> Feedback please. I'm not quite sure what to do about this. It would certainly
> be simpler if it was on-link only (i.e. discovery loop count limited to 1).
> However, the case where there are two discovery responses and only the 2nd one
> supports Rapid Mode is tricky.
> 
> Regards
>    Brian
> 
> _______________________________________________
> Anima-signaling mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/anima-signaling

-- 
---
Toerless Eckert, [e-mail]