Re: [Roll] Benjamin Kaduk's Discuss on draft-ietf-roll-turnon-rfc8138-14: (with DISCUSS and COMMENT)

[Michael Richardson <[e-mail]>]

Martin Duke <[e-mail]> wrote:
    >> Martin Duke <[e-mail]> wrote:
    >> > So I guess the value of the MOP 7 behavior is... to recover the bit
    >> of the
    >> > flags? Fine, I suppose, but this would appear to have two costs:
    >> > 1) If A is still out there when E deploys, this bit would have been
    >> useful,
    >> > but isn't.
    >>
    >> The whole point is that A and E can't co-exist, BY CONSTRUCTION.
    >> The same way that IPv6 nodes can't talk to IPv4 directly.
    >> And we're okay with that, because it's the A/B -> C transition that
    >> worries us.

I'm sorry, the table was hard to read/remember, so maybe I should have said
"A/B can't co-exist with E"

    > I've already lifted the DISCUSS, but I am confused by this response. Yes,
    > in a perfect world all nodes are upgraded to implement this draft before
    > anyone ships anything with MOP7.

We are saying that, not just in a perfect world.
By construction, we saying that was are intentionally making this decision.
We have MOP 5 and MOP 6 still to allocate and possibly a bunch of other stuff
before we get to that place.

Again, IPv6 analogy: state "E" is when we turn off IPv4.
Any node not upgraded is landfill.  Sorry.

    > Is it not accurate that node C could
    > interop with a MOP7-capable node, because C could be a leaf node?

I think, yes.

    > If so,
    > then if A is still lying around then it can't even be a leaf node because
    > there's no way for E to run MOP7 without compression.

correct.
But... this is on 802.15.4 and other links where we use optionally use 8138 compression.
It wouldn't be true at all on 802.11 or ethernet links.
(Such as draft-ietf-anima-autonomic-control-plane's RPL instance)
Some new IP-over-FOO document where FOO != 802.15.4 [or lookalikes], could say
that 8138 was always on, or compression was unnecessary, or ...
Of course, device "A" probably won't have this new fangled as-yet-to-be-invented
interface, so it can't connect at layer-1 anyway.  If A if modular and can
grow such an interface, then I guess it better get a software update too.

    > Maybe it's true that A and B will all be gone, in which case my example is
    > true but irrelevant.


--
Michael Richardson <[e-mail]>   . o O ( IPv6 IøT consulting )
           Sandelman Software Works Inc, Ottawa and Worldwide