[Anima-bootstrap] MB review of Bootstrap charter (was: a repost of summary)

[Michael Richardson <[e-mail]>]

Michael Behringer (mbehring) <[e-mail]> wrote:
    mb> - Probably a nit: I'm missing the words "trust infrastructure". To me,
    mb> we're setting up a trust infrastructure, and we also need to discuss
    mb> the setup of the registrar, CA, not only the bootstrapping
    mb> elements. The word "trust" is completely missing - is there are reason
    mb> for it?

The document does discuss registrar, CA, etc, but we have not been discussing
those in our meeting.   I think that the interface to those things are
relatively well described by current specifications: but there will be
profile work require, I'm sure.

    mb> - Goal 1: Is this the certificate format and fields?

I wasn't sure what "Goal" text you are referring to here, then I realized
that you were referring to the DT Charter text that says:
     >  Goal one: Provide a normative definition for the concept of a
     >  Domain's Identity.

I'm not sure that we have really much work to do in certificate format
and fields.  The IDevID 802.1AR document is woefully underspecified on
what goes into the SubjectAltName, and I think we need to be more precise
here, and maybe that will require a new document.
(such as: https://tools.ietf.org/html/draft-richardson-6tisch-idevid-cert-01
which is entirely stolen from  RFC3779 )


    mb> - Goal 2: I think security needs to also be discussed in the reference
    mb> model, we need to see how we can split that? Or do you see that the
    mb> bootstrap doc would cover that all?

Here is the text from the charter:
     > Goal two: Define the target underlying security model of an autonomic
     > network, in which autonomic devices are enrolled with a local
     > certificate (e.g: IEEE 802.1AR LDevID) identifying the autonomic
     > domain that they belong to by hash of the public key of the trust
     > anchor.

I think that the important part for the reference model is that:
  "Devices have 802.1AR IDevIDs, with replaceable LDevIDs"

Should the reference model/framework include ideas of device resell,
and the problems of supply chain management? If so, then I would love to
have a place to put those considerations.

    mb> - I think there is overlap with the netconf zero-touch work, and we
    mb> should probably be explicit: Do we want to find common groud (would
    mb> hope so), or is it too late? Since that has been discussed, we should
    mb> be explicit about this point I think. Which other groups do you have in
    mb> mind when talking about "adjacent IETF WGs"? Can we list them?

I think that netconf zero-touch work is useful and interesting, but I also
see it as in some ways completely unrelated.  That might be just how I
regard the work, as a kind of uber-secure, NAT and end-user friendly
replacement for DHCPv4+TFTP to get configuration files/boot images.

As such, in the netconf situation, devices show up on (preconfigured)
networks, and call home (for various definitions of "home") to get a
configuration has been determined by mechanisms out of scope.

In ANIMA, devices only call home to establish ownership and to affect
delegation of of trust.  The network that they are "plugged" into may be
hostile, may be unconfigured or mis-configured... Further, the goal of
the bootstrap is not to establish the configuration of the device, but
rather to build a trust relationship with other devices so that each
device can derive it's own configuration.

Having said, this the bootstrap work could be used to provide security to the
zero-touch work, or the netconf zero-touch process could be used to install
the locally significant certificates that bootstrap is about.  Or EST
or 6top or..

    mb> - somewhere the charter should state more explicitly: Define protocols
    mb> and data models for the bootstrap process. I guess that's goal 3, but
    mb> it sounds cryptic. To me, we need 1) requirements, 2) a protocol, 3) a
    mb> data model, including certificate format and fields.

Goal 3 is about how to use the LDevID; this might mean explaining how
it fits into TLS, DTLS, IKEv2, DHCPv6, SEND or some other protocol that the
signaling system might base it's security on.

    mb> - in the reference model I'm referring to the bootstrap process in a
    mb> new way, in the ASA thinking. There is an ASA for a new element, one
    mb> for the proxy, and one for the registrar. (Previously those functions
    mb> were included in the infrastructure, but logically they should be
    mb> ASAs). Do you agree with this thinking, and if so, should we adopt
    mb> this wording here as well?

If you are saying that the proxy and the registrar are functions that the
network should realize that it needs to provide, and therefore should be
described as ASAs, then I agree.   This implies that we have to run the
signaling protocol recursively, and that we have to have more than one
instance of the signaling protocol: one insecure instance, and possibly
multiple secure instances. One shouldn't assume there is a single network
being assembled.. after all, a field can contain many picnics:
      https://www.pinterest.com/pin/216595063299304430/

    mb> - Do we want to include downloading a config? (I read "optimized
    mb> configuration distribution) - somehow that doesn't seem to be in scope,
    mb> in my reading. Can we exclude that? To me, the process ends when the
    mb> device has a domain cert.

Max, can you say more about Goal five?

    mb> - The milestone is very cryptic. Are we planning a new spin of
    mb> draft-pritikin before the IETF? Or is this a completely new draft?
    mb> you say "draft(s)" - what does that mean?

Yes, spin it.

    > Michael

    >> the charter for the WG, and the result is at:
    >> https://tools.ietf.org/wg/anima/trac/wiki/Bootstrap

Charter:
https://tools.ietf.org/wg/anima/trac/wiki/Bootstrap%20Design%20Team%20Charter



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