Re: [Anima] draft-ietf-anima-autonomic-control-plane-06

[Michael Richardson <[e-mail]>]

Toerless Eckert <[e-mail]> wrote:
    >> Not at all. The RPL profile template is at:
    >> https://datatracker.ietf.org/doc/draft-ietf-roll-applicability-template/
    >> (yes, it's expired, never to be published)
    >>
    >> https://datatracker.ietf.org/doc/rfc7733/ (section 4)
    >> https://datatracker.ietf.org/doc/rfc8036/ (section 7)
    >>
    >> are examples of the RPL profile.

    > Great. I thought in Chicago you said you would volunteer converting the current section
    > about Roll profile in ACP spec into that format. Is that offer still
    > valid ?

Yes, I'll still do that, but there are some other BRSKI edits that seem to
take priority :-)

    >> If we aren't going to depend RPI,  we'll need to carefully tweak the RPL
    >> parameters so that we get frequent enough announcements.  We may be able to
    >> leverage the IPsec DPD messages to detect link down's and do reparent events
    >> though.  That should be easily written up in the ACP document.

    > Good point. I guess thats details that would go beyond RPL profile.
    > Text suggestions welcome, otherwise i'll try to make up something.

That's actually exactly what the RPL Profile addresses.

    > Btw: I am very interested to have some more high-available ACP option, but
    > that would be even more work, eg: Something like MRT support with RPL.
    > Maybe we can discuss in Prague..

What is "MRT" in this context?

    >> the IPsec tunnel, it probably looks like:
    >>
    >> IPll ESP RPI IP ULP
    >>
    >> With the IPll ESP RPI part being added/removed at each hop.  Based upon 20
    >> years of building drivers for hardware acceleration, I don't think that this
    >> will matter much to HW accel.

    > Well... If i look at the routing requirements that i see:
    > - ability to have different NOCs are separate roots
    > - ability to have MRTs

    > I think i could easier resolve those requirements with (S/M,D) based
    > forwarding entries than with IPinIP header fields to specify an instance.
    > But i guess that such approaches would be novel for RPL and (see above)
    > therefore require more work.

Different NOCs could create different DODAGs, yes, but it requires InstanceIDs.
There may be some other options.

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