December 02, 2003

Johnson Braves the Mire

I had the distinct priviledge this afternoon of hearing Ralph Johnson (yes that Ralph Johnson) present our hoary chestnut, Big Ball of Mud to his Software Engineering seminar. The mere prospect of this caused me to start reflecting on it...

spaghetti-medium.jpg I still find myself haunted by Kent Beck's critique that this masterpiece of equivocation ought instead to have had the "courage of its convictions".

I've become increasingly receptive to this perspective over the years. Is there something about the metaphors we use to descibe software that blinds us to its fundamental nature? Is that untidy tangle we dismiss as anomalous really part and parcel of building this stuff?

The set of systems that can feasibly be built at all is strickly larger than those that can be built elegantly. For many tasks, and for many teams, this architectural style may be the only possible match.

We made a point of saying, and so it is oft said, that this paper should not be seen as an endorsement of Big Balls of Mud. Indeed, we made it clear that we were, by no means, recommending these kinds of designs. At the same time, we resisted considerable pressure to utterly repudiate this architecture, for a variety reasons.

For one thing, tangled legacy systems leave programmers no choice but to cope as well as they can. Infantry-style teams and processes insure system made in the team's image. Some problem domains may pose requirements so inherently muddled that must inevitably be mirrored in any possible solution.

While I still can't say I recommend this approach, I'm convinced that it was high time that someone try to describe and explain it. This architecture may as well be placed honestly and openly on-the-table, since its spectre looms large over its more respectable alternatives.

Noble et al. have nominated this work as perhaps one of Computer Science's first post-modern works. I'm quite sanguine about this characterization. I realized during Ralph's lecture that a legitimate descontruction of our argument, that is to say, a characterization of our unstated posture, might be that small teams of skilled chraftsmen can beat an underskilled human wave approach every time. Were we really advocating a world made in our (presumed) image?

And, this Slashdot nugget on role fragmentation suggests yet another possible perspective on this multi-faceted issue. (And, yes, it feels rather pointless to echo a Slashdot post in one's own weblog, but what the hell...)

Ralph observed at one point that Big Balls of Mud are what you get when you throw an army of Visual Basic programmers at a problem, and further, that that's just what you ought to expect to get. It's yet another corrolary to Conway's Law. Given that you've elected to employ a large number of modestly talented "infantry level" coders, a haphazzard hodge-podge is what you should properly expect.

Now, back to Beck's challenge. Kent made this remark, I've recently realized, back when his ideas about Extreme Programming were taking shape. XP would ultimately relegate concerns about design aesthetics to a secondary, or even tertiary position in its pantheon. You Are Not Going to Need It demanded that design flourishes be eschewed in lieu of immediate, established requirements. It is a defiantly utilitarian process that ultimately came to scoff at the petty egotistical inclinations of designers towards generality, extensibility, and elegance. Indeed, it gently mocks these inclinations as soft of wasteful hubris. Looking back, it seems that Kent had begun to see the Sirens of Elegance, of High Road Architecture, as of the major obstacles on the road to a more dependable software development process.

Posted by foote at December 2, 2003 04:56 PM