Re: [dwm] jr-patchset up to date

On Wed, 30 Aug 2006 11:34:05 +0200
"Anselm R. Garbe" <arg_AT_10kloc.org> wrote:

> I'm also going to generalize the 10kloc project into 'shortest code
> project' - which will only accept software <10kloc, but with somewhat
> more liberal statements.

I've been meaning to comment on that...

I wholeheartedly agree that the less loc a program has the less errors
it has, and that people should try to write less and not more. Late E.
W. Dijkstra said it well: '[I]f we wish to count lines of code, we
should not regard them as "lines produced" but as "lines spent"'. [1]
(As an aside, he was also of the opinion that the shortest solution is
usually the most elegant and the most efficient one.)

However, I'm not sure about a fixed numerical limit, because the limit
seems arbitrary. Why 10k? Why not 11k or 9k? Although the limit is good
for big projects in that there needs to be effort to reduce the loc,
isn't the opposite true for smaller projects? I can easily make a 1kloc
Hello World application, and it can hardly be said to fit this
philosophy.

I'd rather see some other metric in use, something that binds the
abstract size of the design to the concrete size of the implementation.
Yes, I know loc is easy to count, and it's probably why it has been so
popular since 1960s. I also understand that creating a new metric or
applying something like COCOMO isn't that appealing, either.

It would also be good to make it clear on the web site that if you are
able to remove one loc and at the same time increase the level of
obfuscation, it's against the philosophy (right?).

[1]
http://www.cs.utexas.edu/users/EWD/transcriptions/EWD10xx/EWD1036.html

-- 
Ville Koskinen