Re: [Gen-art] Gen-ART review of draft-ietf-idr-rfd-usable-02

[Randy Bush <[e-mail]>]

> As the document shepherd this same phrase struck me a different way so
> I let it go forward.
> 
> "This document is not a panacea, nor is it a deep and
>  thorough approach to flap reduction."
> 
> My understanding is that some authors (researchers and engineers) were
> from Japan.

well, gaijin living and working in japan for some years.  but yes, if
our natural researchers' instincts were not cautious/precise enough, we 
probably have been even more infected by this culture. :)

> As careful engineers their concern is to be precise on the work they
> have done.

precisely :)

> They have addressed immediate concerns and known issues as documented
> (2 years of work), but they feel additional work can be done.

take a look, for example, at geoff huston's work in the area.  it is
algorithmically more appealing, looking at the fundamental properties of
flap, as opposed to our continuing rfd's statistical approach.  but in
actual implementation, it turned out to be no more effective than our
hack, was vastly more costly in cpu and ram in the router, and therefore
was abandoned.

our hack merely tunes existing constants in existing algorithms and
running code.  we get a non-trivial win really cheaply.  but it would
take amazing hubris to call it a fundamental approach to flap reduction.

randy