[Arcana] the travelling salesman problem solved!

Jim Blandy jimb at red-bean.com
Fri Mar 21 19:06:40 CDT 2008


Matthias Felleisen often took similar positions to the ones Dijkstra
takes in that essay.  They're full of crap.

The reason software matters in society, instead of being an obscure
branch of mathematics or engineering, is that people do astonishing
things with it, like Smalltalk, the Mac, the Internet, IRC, the
world-wide web, video games, and cell phones.  All of those things
were, by Dijkstra's standards, ad-hoc slap-dash attempts to just get
something that works at all, perhaps refined later in the presence of
much more money.  The techniques Dijkstra claims one is a moron if one
does not use increase the investment required to get something that
works so much that these things would have come decades later, or, in
the case of the skunkworks projcts, wouldn't have happened at all.

Dijkstra's correct, I think, that it's silly to expect flawless
software without the help of formalism like those he's promoting.  But
correct software five years later is demonstrably not valuable
compared to shockingly cool software now.

That is, I think Dijkstra is correct that his techniques are necessary
to achieve the goal he has in mind, but I think that Dijkstra's goal
is not socially or economically interesting.

(I don't know if this is a question it's worthwhile to have an opinion
on, but I've got one, dammit.)




More information about the Arcana mailing list