[e2e] Open the floodgate

David P. Reed dpreed at reed.com
Wed Apr 21 14:23:09 PDT 2004


At 01:44 PM 4/21/2004, Cannara wrote:
>So, nowadays, the lack of systems-management discipline in The Internet leaves
>us with few options.

This language is quite interesting.   To readers of George Lakoff, like 
myself, it's clear that the unexamined assumption is that "discipline" is 
inherently good.   A world with strict rules, a stern "father", etc. 
eliminates the possibility of chaos, replacing it with a utopia where every 
engineer is given a spec, and delivers precise implementations, or every 
design is based on thorough knowledge of the operating regime of the target 
system, so it cannot fail - the only possibility is that the specification 
doesn't match the inputs.

Mussolini impressed the British and Americans with his delivery of 
discipline (making the trains run on time) to the world.   There is no 
dispute that he was disciplined and professional.   The problem was that he 
idealized the wrong state of the world as the most desirable.

Given a choice between discipline of design and usefulness it's clear that 
the latter is more important.  The former is merely a tool, not a 
goal.   As such, the limits to the utility of discipline need to be understood.

A slavish adherence to discipline can easily lead to failure - and often 
does.   The most disciplined and beautiful technical project in memory is 
the Iridium satellite network.   They delivered a ten-year, billion dollar 
project, to spec and on time.   Every technical requirement was met, and 
wonderful breakthrough invention was completed on time.   But anyone 
looking at the project from the outside in the bigger picture can see that 
it was an incredibly costly failure *of engineering*.   Could it have been 
different?   Why yes - during the 10 years of its development there were 
many times when its design could have been adapted to incorporate new 
information about customer needs and requirements, about competitive 
alternative solutions, etc.   A less disciplined approach might have 
avoided marching that brilliant army in perfect formation off of a cliff - 
or even as the first part of the army marched off of the cliff, the rest 
could have decided to save the bulk of the investment by turning away and 
redeploying their resources against any number of alternative objectives.

In contrast, the Internet continues to be successful, continues to adapt to 
new requirements, and continues to incorporate massive sets of innovations.

Anyone who teaches engineering as if the only thing that matters is 
discipline is not teaching engineering.   Engineering as a profession is 
not only about predictable delivery of a well-specified solution on 
time.   That is what a bricklayer or carpenter or contract programmer 
does.  It's a skill or a craft - the mastery of a tool.  But craft 
discipline is not engineering anymore than playing the precise notes 
written on a piece of paper precisely on the beat is performing music.

Sure, it's always important to pay attention to discipline.   But the 
Internet seems to have sustained an appropriately effective level of 
discipline in its development - certainly adequate to the task, while also 
sustaining a crucially important level of exploratory and experimental 
development that has enabled a wide variety of unanticipated innovations to 
emerge - most of which would not have emerged from any "disciplined" 
process of the extreme sort that optimizes for the past while marching 
backwards into the future. 



More information about the end2end-interest mailing list