[Server-sky] Server Sky and Good Questions

[Keith Lofstrom]

On Fri, Jun 21, 2013 at 01:20:26PM -0700, Jonathan "Duke" Leto wrote:
> Bring your amazing ideas and passion in spades, but leave negativity
> at the door.

On Fri, Jun 21, 2013 at 10:43:04PM +0200, Eugen Leitl wrote:
> There's no negativity -- certainly not from me.

Asking the obvious questions may be quite uncomfortable but it
is not negativity (which is usually perjorative).  We will hear
the same questions thousands of times before this gets built,
and we must learn to answer them with candor, in a thought
provoking way.  And we must carefully gauge the reaction of our
listeners, because we will learn better ways to answer from that.

We must keep our questioners engaged, because we will eventually
hear the hard questions, the ones we have never heard before and
that we do not have answers for.  One question like that, followed
by research and design, can save tens of millions of dollars at
implementation time.  My main contribution to most projects I have
worked on is finding the questions that haven't been asked yet.

So I appreciate amazing ideas, and passion, and loyalty, and
friendship.  That will get us through the tough times.  But it is
better to seek out the tough times, the probing audiences, even
the reflexively negative, because when we have answers that win
over that crowd, we are halfway to hard-earned success.

A huge part of server sky will be software, and with the right
systems in place we can fix much of that after deployment.  But
the hardware will be literally irretrievable,  and the path 
towards hardware that has a good chance of working is extreme
self-skepticism, nights without sleep, and not a little hollering.

Take a look at the excellent JPL videos on the development of the
recently landed Mars Science Laboratory.  One of my heros, Adam
Stelzer, led the entry and landing team, and his continuous searching
for flaws shows in those videos, and in the successful result.  My
wife, watching some of the videos with me, was struck with how these
people were just nit-picking as I am.  You've gotta be that way in
semiconductors, or that mask set you just spent 5 million dollars 
making will generate useless silicon after weeks in the fab.  

As a former boss sagely observed, "too many oughta works is a not
oughta work."  We keep looking for flaws until we stop finding them,
and after we've designed in every tool for finding and patching 
around the inevitable remaining flaws during production and deployment.
And we will keep developing and improving too fast to have the luxury
of removing our patch tools.

So, thanks for intelligent questions.  We have answers for the ones
asked so far, but those answers need improvement, and I am grateful
for the opportunity to participate in finding them.

Keith

-- 
Keith Lofstrom          keithl at keithl.com         Voice (503)-520-1993