Re: I-D Action: draft-troan-6man-p2p-ethernet-00.txt

[Michael Richardson <[e-mail]>]

Brian E Carpenter <[e-mail]> wrote:
    > An interesting draft on an important topic, but I think it
    > oversimplifies the problem in (at least) two ways:

    > 1. Even if we had new improved switches and hosts, they would for many
    > years be mixed with unimproved switches and hosts on the old (emulated
    > yellow cable) model. It's already bad enough when you mix switches that
    > do MLD with dumb switches that really do only emulate yellow cable.

I've experienced a switch with broken MLD recently, and it breaks everything
other than IPv6-LL communication... but IPv4 (which uses broadcast) continues
to work!

I think that any well constructed RFC8504 host, given an RA with L=0, will hairpin
all traffic to the router.  
That effectively removes all the multicast dependancies.  This is what all
wifi routers should do, in my opinion.
I think we don't have to have any flag days in theory.
In practice, we will probably find bugs with L=0 on some OSes.

    > 2. I think we should be expect mesh rather than hub and spoke in the
    > quite near future.

    > Which all takes us back to homenet, doesn't it?

Yes, kinda, but kinda not.
I think it could be much simpler for some things.

-- 
]               Never tell me the odds!                 | ipv6 mesh networks [ 
]   Michael Richardson, Sandelman Software Works        | network architect  [ 
]     [e-mail]  http://www.sandelman.ca/        |   ruby on rails    [