Large systems (million SLOC plus) are often a reflection of adding many controls in an ad-hock fashion. Because removing the controls breaks an “intersection” in some component or subsystem, people build new “intersections” in the system, sometimes right on top of the old one. So these systems are typically 2-10x times the size they probably could/should be. Now the million+ SLOC system is really an obese 100k-500k SLOC system.

I long for the days of Python development where the toughest thing I’d run into would be a really nifty list comprehension. Where it was easier to write the test first than not to. Could you build really large systems with Python? I’m certain of it, even though I’ve never done it. Perhaps that’s the best possible answer!

On woonerf (itself a woonerful word): I wonder whether my urban neighborhood would be better without all the STOP signs, but with enforced crosswalks. I think a problem we have that other places don’t is that a lot of our traffic is out-of-town commuters, who don’t have any attachment to our community, and whose cars I will key the next time they stop in the crosswalk while I’m pushing a stroller.

I would not know how that would help your analogy though.

I wrote about the same ideas a while back here (http://www.itworld.com/nls_ebiznaked050426). I did not know the Dutch name at the time : woonerf. I like it!