Subject: Re: rocket science
From: Ben Tilly <>
Date: Mon, 28 Feb 2005 13:08:19 -0800

On Mon, 28 Feb 2005 11:19:08 -0800 (PST), Tom Lord <> wrote:
>    From: "Forrest J. Cavalier III" <>
>    [The examples reported on RISKS have not yet caused wide-scale
>     economic disruption you think they never will.  Business
>     is concerned with real things, not imagined risks]
> Please consider the example of the computing systems that run
> financial markets and the rise of intelligent consumerism
> in the markets that purchase those systems.

I've considered the example.  I've also concluded that you've
probably never actually worked in finance.

> In those markets, the risks I cite are taken quite seriously
> and dominate spending decisions.

There's a lot of CYA in finance, true.

Then people proceed to work in an amazingly haphazard way.

Obviously this varies widely between organizations.  But there
is a definite tendancy to throw lots of hardware and money at
problems before taking what I'd consider basic steps, like
using source control consistently.

The problem is, of course, that quality processes are hard for
consumers to measure.  (Furthermore most of us think that
we're better than we are.)  Because it is hard to measure it,
people do not generally make purchase decisions based on
it.  Instead they decide based on other things that they think
might be related.  (Like the quality of the advertising

> My claim here is that that's part of a trend:  comperable
> levels of seriousness will spread to other segments of
> the computing system industry as consumers recognize that
> so much of what they value depends critically on computing
> systems.

Yup.  But I'm taking that to mean something different than
what you probably intended...

> Source code and development process quality metrics will therefore
> emerge from the esoteric labs of NASA and NASDAQ and become more
> mainstream, especially in enterprise-level purchases.

Nope.  The industry will continue to engage in "satisficing
behaviour".  And the bar will prove to be at different levels
for different applications.

> Proprietary developers are currently in an advantaged position to
> prepare for that spread of a concern for quality.

My experience says that free software is usually better.
However extreme quality is something that free software
has a hard time delivering.  (Though OpenBSD shows
that we can do better.)

> Free software developers are currently disadvantaged because
> social and structural issues present significant obstacles to
> improving development processes and because some of the software
> architectures we have collectively chosen have intrinsic quality
> problems.

Have you been listening to the capabilities people? :-/

> FSB leaders ought to be thinking and then acting, these days,
> to overcome their disadvantages and overtake the advantages
> of proprietary developers in these areas.
> This will be a difficult task because it appears it will require
> inter-firm cooperation in areas where it has not previously
> taken root.  Generally speaking, FSB firms need to learn to cooperate
> more on software development issues that can be described as "management
> issues" as contrasted with "day-to-day hacker issues".

Won't happen.  Sorry.