Re: [v6ops] comments on draft-xiao-v6ops-nd-deployment-guidelines

[Michael Richardson <[e-mail]>]

Xipengxiao <[e-mail]> wrote:
    >> I think that the better approach is to have a section two ask
    >> questions (i.e. a self-assessment) as to what kind of scenario they
    >> want to do.  Some of the results might be aspirational vs actual.  As
    >> in, "I really want to get to UPPH support, but not all my hosts are
    >> capable yet, so until then, I want MLSN"

    > [XX] I read your message 3 times, but I still don't fully understand
    > the logic flow here.  The current logic of the draft is (1) review the
    > known issues and solutions, in Section 2.1-2.4 (2) analyze the essence
    > of each solution to see what mechanisms make it effective, to get some
    > insight, in Section 2.5 (3) apply the insight to future deployment
    > scenarios, and discuss the applicability, in Section 3.1 (4) summarize
    > the whole practice into 4 guidelines for future ND deployers to
    > consider, in Section 3.2.  Maybe this logic in my mind gets in the way
    > of understanding your logic.  So if you can further elaborate your
    > logic I will appreciate it.

This is a very good structure for an academic review paper.
It puts all sorts of information and background up front.
It is the mathematical way: write down assumptions, derive some lemmas from
that, work through some proofs and generate some conclusions.

In general, few operators read academic papers.
Evidence is that scientists don't really discover things in the way that math
papers and textbooks suggest.  Rather, they start from the goals ("the
applicability") and then work backwards to what evidence they need to support
that hypothesis. (Then do experiments)

So I'm suggesting that a more useful opertional document would start by
describing different kinds of target opertional situations, what the
characteristics of the desired networks are.  Then determine what mechanisms
would be support those goals, and then deal with what the limitations might
be.

    >> The Security Considerations will need to do a lot more.

    > [XX] This draft only summarizes existing solutions, and provide
    > guidelines on how to select them based on the usage scenarios. It
    > introduces no new solution.  So I think it's fair for us to say it does
    > not introduces new security issues.  What else do you think are needed?

If an operator picks solution X vs solution Y, then they need to understand
what X does not do, that Y does do.

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