Re: [Roll] multicast DAO

["Julien Abeille (jabeille)" <jabeille at cisco.com>]

Hi Phil, 

Sorry for the delayed response. 
> -----Original Message-----
> From: Philip Levis [mailto:pal at cs.stanford.edu] 
> Sent: vendredi 20 novembre 2009 19:31
> To: Julien Abeille (jabeille)
> Cc: ROLL WG
> Subject: Re: [Roll] multicast DAO
> 
> 
> On Oct 28, 2009, at 9:29 AM, Julien Abeille (jabeille) wrote:
> 
> > Hi all,
> >
> > regarding usage of multicast DAO: in the following scenario:
> >   A
> >  /  \
> > B  C
> >
> > if the purpose is to to allow B and C to talk to each other 
> directly, 
> > the multicast DAO is equivalent to an unsolicited NA.
> >
> > In the scenarios like:
> >
> >    A
> >   /  \
> > B   C
> >       |
> >       D
> > if the purpose is to have B talk to D through C (instead of 
> A then C), 
> > then the multicast DAO (including destination prefix option) makes 
> > sense.
> > Is this scenario needed by some applications?
> 
> Julien,
> 
> The text in -04 reads:
> 
> "The objective [of multicast DAOs] is to enable direct P2P 
> communication, between destinations directly supported by 
> neighboring nodes, without needing the RPL routing structure 
> to relay the packets."
> 
> So the case it's concerned with is the latter one: B to C to D.  
> Clearly one could use NA for the B-C case.
> 
Two questions: is it really a scenario (B to C to D) that happens a lot.
I feel it is not generic at all, hence we should focus on a more generic
way to solve the P2P problem. 
> Point-to-point communication between nodes is needed by some 
> applications. Section 5.2.2 of 
> draft-ietf-roll-building-routing-reqs-07:
> 
> "A network device MUST be able to communicate in a 
> point-to-point manner with any other device on the network."
> 
Sure, and DAOs (whether stored only at root are not) solve this issue.
May not be the most efficient, but the sceanrios multicast DAOs solve
are so limited that we will end up with 10 solutions for 10 scenarios if
we go that path.

I would personaly drop multicast DAOs

Best,
Julien
> Do you agree? Is it OK to resolve this ticket?
> 
> Phil
> 
>