Re: [Roll] [roll] Discussion around the design of draft-ietf-roll-nsa-extension based on MRHOF

[Michael Richardson <[e-mail]>]

Remous-Aris Koutsiamanis <[e-mail]> wrote:
    > After addressing all of Alvaro's and IANA's latest comments, the issue
    > of the Objective Function design came up.

    > Currently, the Common Ancestor Objective Function (CAOF) is "based on"
    > MRHOF. "Based on" in the sense that it is described in terms of a diff
    > to MRHOF.  The CAOF defines the concept of selecting a secondary
    > "Alternative" Parent, in addition to the - primary - Preferred Parent,
    > to be used to increase the reliability of the upstream transmissions.
    > Since the CAOF defines a complete objective function, it must explain
    > how the preferred parent selection works. This is the part that is
    > based on/inherited from MRHOF.

Thank you for this email, as I just didn't understand the discussion during
today's session.

    > However, the selection of MRHOF as a base has come into question.

    > The two reasons given were: 1. We can avoid basing CAOF on MRHOF, and
    > thus make it more generic, by simply explaining how the CAOF can modify
    > *any* objective function, instead of just MRHOF, to add the ability to
    > select and use an alternative parent.

If you can do this: explain how it can modify any objective function, then
let's do that.

    > 2. According to Pascal Thubert,
    > MRHOF has limited adoption in the industry, and thus basing CAOF on
    > MRHOF may limit its adoption.

This implies that in the industry, there is, alas, very very little
interoperation :-)

My implementation, btw, has a NOP for the objective function, I seem to
recall.  I just never got to writing it.  What's really funny, is that I
started the implementation in order to better understand that part, but in
order to get there, one needed the rest of the structure.

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