Re: I-D Action: draft-carpenter-6man-lap-00.txt

[Michael Richardson <[e-mail]>]

Brian E Carpenter <[e-mail]> wrote:
    > Two names, adding Mark Smith's SAID suggestion. And yes, having a name
    > or names that express that there's a decision involved (is this
    > acceptable?)  is IMHO a way to track progress in the discussion as well
    > as in deployment.  It's a very mild form of a thinking tool (as per
    > Daniel Kahneman).

I'm always in favour of having names to talk about things.

    > If somebody wants to deploy subnets with a maximum of (say) 256 IoT
    > nodes, and doesn't care about entropy in the IIDs because of privacy
    > aspects of the specific scenario, why should we care if they choose a
    > SAID of 8 bits?

    > (Why might that be a good thing? Because it would allow you to express
    > addresses within the subnet in 8 bits. For a constrained IoT node that
    > might be very important.)

Just to add, that this is real, not just "might", but that it's 16-bits.

802.15.4 networks support 64-bit EUIs for layer-2, as well as having a 2-byte
version.   Assigning the IIDs to match the L2 ID results in significant
compression savings.    The 64-bit EUIs can be manufacturer assigned, or
could be randomly generated.  The 2-byte addresses can be assigned via
a variety of ways, DHCPv6 and draft-ietf-6tisch-minimal-security come to
mind.

Today, IoT devices do not visit Starbucks or Airports and worry about being
tracked.  In the future when we have actual PANs of devices on one's person,
it is my contention that they won't talk to the Airport WIFI either.
(Likely they will be ULAs, with VPNs back to places that care about them)
However, even when encrypted, 802.15.4 networks leak L2-addresses just like
wifi networks do, but 16-bit layer-2 addresses are likely meaningless.

In 6tisch/6lo 802.15.4 networks, the network is still /64, there are just lot
of zeroes.

    > That's the kind of discussion we can have with such terminology.  If we
    > can't have that discussion, we're throwing away valuable flexibility
    > for the universal deployment of IPv6.

    > I don't think anybody is arguing against LAP=64, SAID=64 as the general
    > default.

+1.

--
Michael Richardson <[e-mail]>, Sandelman Software Works
 -= IPv6 IoT consulting =-