RE: improving WG operation

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hallam-Baker, Phillip said:
>
>> for some, "let the market decide" is a religious statement.  it's
>> generally based on an unexamined faith in market conditions as an
>> effective way of making a good choice among competing technologies.
>
> I don't accept the ideological case for or against free markets.
>
> My point here is limited to the case in which a working group is unable
> to come to consensus over two disjoint proposals and each group refuses
> to compromise with the other.
>
> Agreement would certainly be the best outcome in the case where the
> differences are due to personality issues. But there are also cases
> where one proposal is in fact distinctly inferior, usually because the
> adherents are bought into some obsolete dogma or other.
>
> For example there is no way to negotiate a compromise between die hard
> adherents to the end-to-end security primciple and proposnents of an
> edge based security system. The two architectural views are entirely
> incompatible and cannot possibly be reconciled.

actually I disagree with you on that point (I agree with most of the rest
of the above).   there is at least sometimes a role for perimeter-based
security in addition to end-to-end security.  the trick is to understand
the strengths and weaknesses of each as they apply to a realistic threat
model, not to assume a priori that either one can exclusively do the job. 
in my experience the disagreement between die hard adherents usually
amounts to an underlying disagreement about the threat model - where both
sides may have oversimplified it.

> What I have observed in these divisions is that it is actually quite
> rare to have two factions of implementers. What is much more common is
> that you have a group of folk who are building something and another
> group of rock throwers who won't build much more than a bunch of
> prototype code that only worksw with their own system.

yes, but often this is because it's much easier to build something that
implements a naively simple version of a protocol than to design and
implement a protocol that will actually work well in a realistic range of
scenarios that will be encountered in wide-scale deployment.  that's why
"running code" by itself isn't worth much anymore.

> In other words letting the market decide comes down to who has the best
> code and the best deployment strategy.

depends on what you mean by "best".  if you mean the strategy that gets a
lousy product out into the market in the shortest amount of time and
attempts to lock in customers, you're right.  if you mean the strategy
that provides the most long-term benefit to the community, you're wrong. 
there's no substitute for engineering.

we would never consider building a bridge, building, ship or large
aircraft, without careful understanding of the problem to be solved,
multiple design/analysis/feedback cycles, etc. the investment in these is
quite obviously so great that we want to minimize the potential to invest
that much in a poor design.  and yet protocol designers will happly invest
similar sums - or even more - of their customers' money in poor designs.

of course, sooner or later the market will probably figure out that
investing money in half-baked protocols or implementations of those
protocols is a poor idea.  the market does learn, it just takes so long to
do so that its errors are very expensive for everyone.

Keith


_______________________________________________

Ietf@xxxxxxxx
https://www1.ietf.org/mailman/listinfo/ietf


[Index of Archives]     [IETF Annoucements]     [IETF]     [IP Storage]     [Yosemite News]     [Linux SCTP]     [Linux Newbies]     [Fedora Users]