The Underappreciated Art of Duct Tape Programming

Three Kings

In an earlier, more naïve era of my career, I had three software development “heroes”: Jamie Zawinski, John Carmack and Joel Spolsky.

Jamie Zawinski grabbed my attention because he was at the center of the internet revolution, right in the bowels of Netscape from the outset, and had established a pattern of posting surprisingly pragmatic comments that defied convention.

It was extraordinary to read someone openly critical of their own organization, especially without it being retracted or redacted the sobered-up or calmed-down next day, and where the author didn’t hide behind anonymity.

Jamie let us commoners see the sometimes ugly mechanics behind the curtain. He also revealed a very interesting workplace that was foreign to the gray-walled cube world that most of us lived in.

Georgetown Fall Fair

This was at a time when Microsoft really had almost unthinkable dominance over the industry, so to hear Jamie discuss the travails of cross-platform development was like going out of bounds at a tourist resort. Seeing what the brochure didn’t show you.

An SGI box IRIX How exotic!

Another of the kings, John Carmack, was blessed with “F-you money” from the incredible success of some of his earlier projects, along with a proven abundance of intelligence and skills, so he too had the luxury of entertaining a surprisingly realistic and pragmatic perspective. He was a principal driver of the evolution of GPUs and gaming hardware, and you can owe thanks for some of the capabilities of your console or dual-GPU rig to his desire to make shooting things in first-person shooters hyper realistic.

Carmack was also one of the original “bloggers”, regularly posting lengthy “blog entries” by repurposing the UNIX finger facilities.

Joel Spolsky is a bit of the odd-man out in this trio. While he did have the requisite first-initial, he wasn’t known for extraordinary technical acumen (beyond having worked on Excel in some earlier life), but hear me out, please.

Joel ascended to Kingship – at least in my personal hierarchy of industry royalty – just after the dot com crash, when CMM factory-line initiatives started to become the mythical silver-bullet: This was an era awash with articles gushing about the amazing adoption of CMM5 among offshoring firms.

Many organizations were striving to reduce software development to an assembly line of easily interchangeable cogs, both of code and people, achieving a utopia where the process would become perfectly predictable and repeatable if only you filed enough forms.

Joel spoke up for developers when most were absurdly blaming the .COM collapse on dual-monitors, Aeron chairs, and inflated developer egos, as if taking developers down a notch and having them sit on a cold rock would have made selling kitty litter online a good idea.

He was essentially an enlightened pointy-hair blogger, and while I wouldn’t look to his blog for technical advice (Wasabi!), he really understood developers and the process of getting software built. And he was willing to risk his own nest egg and put his money where his mouth is, having since built a reasonably successful company in Manhattan that most of us should be envious of.

Unbound by Convention

What made these three really stand apart in a sea of cheap advice-givers and pundits was that none of them were writing to get a job or even necessarily to keep one. Joel made his own bank while the other two were of such technical esteem that they had little to worry about professionally regardless of what they might say.

They weren’t coerced into railing off the latest buzzwords and best practices, deferring to the latest silver-bullet best practice pattern-based UML diagramming system 3NF data warehousing factory built on a n-tier service-oriented, aspect-oriented, polymorphic framework so that they can get the approving nods from the nervous
masses and clueless PHBs.

They didn’t worry about offending a boss who held some sacred cow that if only you did it the way she read on some best practices blog, everything would be fabulous, at least until that initiative fails and you move on to the next cure-all.

The three kings were just saying it like they saw it, which was and still is rare.

Eventually Joel ran out of things to talk about and switched his blog to mechanically regurgitated repeats; Carmack got lost endlessly perfecting the noble quest of simulating head shots when he wasn’t reaching for the stars; and Zawinski decided to engage in endless battles with the city of San Francisco over his money sink of a late-night dance club (if you read his blog about DNA Lounge early on, you could almost smell the contractors taking this dotCOMinaire for a ride).

Maligning Metaphors

I was delighted to see Joel return from effective blogging retirement, and my enthusiasm exploding when I saw that it was a post about Zawinski!

A royal duet!

Okay really Joel was selling a book – like his partner-in-crime Atwood, he seems to be motivated to post by Amazon Affiliate bucks these days, credibility undermined by that kickback – however he chose the Zawinski chapter as his pitch, talking admiringly about how practical and “get ‘er done” (paraphrased) jwz was about his craft, doing so with a present tense that betrays a certain blissful ignorance about Jamie’s career path since.

Joel labeled Jamie the “Duct Tape Programmer”, which was a description that Jamie took as “damning with faint praise”. Joel has long been against architectural astronauts, so he seemed to excitedly hold up Jamie as the successful counterpoint.

Perhaps “duct tape” is a bit of a metaphorical overreach, causing many to envision some Tim the Toolman ‘Ar ar ar’ hack.  Pragmatic or practical probably would have been more accurate, though it would have made for a less contentious entry.

Never mind that Jamie worked within extremely tight timelines, using technology far less advanced than what we have now.

Joel’s entry raged across the social news sites, with the regular suspects popping out of the woodwork to declare it a grievous offense to all that is all that is good in the world of software development. Lots of blog replies parroting the standard best practices appeared, their authors clearly hoping that their boss and any future employers will see how studious and diligent of worker bees they are.

Who Decides on Best Practices?

The people who are the most certain about software development patterns, practices, and technologies are usually the people who have the least reason to have such certainty.

I’m going to be a bit trollish while I go to the extreme and say that many of the oft-quoted leaders of the field, responsible for much of the unquestioned wisdom-bites, have little to demonstrate why they’re in a position to preach.

The revered Fred Brooks, author of the Mythical Man-Month, came into a position of considerable influence largely by leading a project that was by most accounts a massive failure. That would be fine if there was but one way to fail and he found it for the benefit of all, but there are an infinite number of ways that a project can fail.

Of course you must learn from failures, but my experience has been that the explanations for failure are often a worse-than-useless distractionary tactic: When a team technically fails to accomplish what they set out to do, expect the post-mortem to be full of nonsensical misdirection about how everyone and everything else is to blame.

How many post-mortems include the statement “I grossly overestimated my own capabilities” I suspect few.

Steve McConnell is another well-known author in the field, revered for his software development books (though many strangely overlook After the Gold Rush, where McConnell knee-jerk responds to the dot COM collapse by advocating an ill-considered licensing system for software developers), but his professional experience seems to be limited to working on TrueType at Microsoft, and some nebulous software development at Boeing, after which he took on the role of telling the world how software should be developed. Now he consults with pointy-hair bosses to unknown outcomes.

Don’t get me wrong, I have both of them in the bookshelf behind me, and read and greatly enjoy their opinions (Brooks’ observation about second systems is more profound and important, I think, that the over-quoted man-month snippet), but really, let’s keep some perspective and stop using it like they’re the incontestable word of truth.

I read them critically and with an open mind, not taking it as the voice from an all knowing deity, but instead the perspectives of a couple of guys drawing from their

Georgetown Fall Fair

Of course, the esteemed Fred Brooks and Steve McConnell exist in a realm far above most silver-bullet cheerleaders in our industry. These successful authors actually
dirtied their hands with actual software development, refactoring their opinions over the years into refined perspectives. I select them merely as “absolute best case” examples.

More commonly the people who most ardently advocate certain practices and approaches have achieved little, usually having nothing to back their conviction but self-interest and a desire to look like they know what they’re talking about, having associated their id with “correct” approaches.

They just clutch onto whatever they hear is proper and start repeating it like a novelty birthday card repeatedly opened. They’ll tell you that should develop like an ecommerce site, despite not being an ecommerce site; like you’re NASA, despite not being NASA; like you make the software for a pacemaker, despite actually making an ebay auction sniping tool.

Why do I hear the word “pattern” from mediocre or non-developers more than I hear about it from experienced developers, always stated as some sort of conclusive statement?

Why do we accept that a chimp-level of software development skill is acceptable for maintenance programmers, capable of understanding only the most infantile code that is carefully decorated with “Coding for Dummies” comments?

Why is “We should use UML” the desperate last-ditch fallback of failing teams everywhere?

Unit testing, or the more early-loaded TDD, can be great, but it isn’t a panacea and is an extremely poor substitute for actual craftsmanship.

Moving beyond the non-developers giving their unwanted opinion on how software should be built, the other class of destructive noise is the advocacy of silver-bullet methodologies during the honeymoon period.

Great, you built a sample app on RoR/Haskell/Scheme/python or whatever else is the cure-all platform that profoundly changed your world view.

Here’s a nickel. Go build a real app then tell us how revolutionary it is now. I don’t discount the advantages, but advocacy based upon toying around is of little use to real projects. Extrapolating it up is foolhardy.

Oh look, another guy telling us how switching to the Dvorak keyboard layout made him regular and makes his code smell like cinnamon. Here’s someone saying that they slept 4 hours a night by taking 20 minute catnaps, proven out over their two day sample period. This guy says that having a 400×200 single-app screen on a netbook made him a perfectly focused coder. Here’s a dieter who is certain that they’re onto an incredible, beats-the-laws-of-thermodynamics diet now that they’ve followed it for a whole six hours.

The Emperor Has No Clothes!

The fairy tale “The Emperor’s New Clothes” has significant relevance to the software development field. To quote the plot summary from Wikipedia

An emperor of a prosperous city who cares more about clothes than military pursuits or entertainment hires two swindlers who promise him the finest suit of clothes from the most beautiful cloth. This cloth, they tell him, is invisible to anyone who was either stupid or unfit for his position. The Emperor cannot see the (non-existent) cloth, but pretends that he can for fear of appearing stupid; his ministers do the same. When the swindlers report that the suit is finished, they dress him in mime. The Emperor then goes on a procession through the capital showing off his new “clothes”. During the course of the procession, a small child cries out, “the emperor is naked!” The crowd realizes the child is telling the truth. The Emperor, however, holds his head high and continues the procession.

Too often the software development industry suffers for lack of someone crying out. We often just go along with it, listening to the declarations of non-developers and maintenance programmers as if they speak unquestionable truth, all while discarding any counterexamples as mere aberration (“Well not every team has superstars you know! We aren’t all John Carmack!”).

Everyone withholds contrary, pragmatic “Well it isn’t quite so cut and dry…” opinion lest they look like a “hack” to a present or future employer or nervous, cargo-cult embracing peers, smiling politely while the never-coded, overconfident guy acronym drops about things he doesn’t understand in the daily stand-up.

The more you know, the more you’ve experienced, the less obvious the world becomes, and the more hesitation before offering up opinions. The less ease there is to criticize the path of others when it has yielded obvious success.

Opinions come quickly to experts and morons. Few of us are experts.

Jamie Zawinski had unique conditions under which he unquestionably succeeded. Many, with the seeming clarity of hindsight and the ability to project whatever imaginary timeline one desires, will look back and comment on how the codebase got rewritten, purportedly twice, and how eventually the product was squashed, stomped out of relevance by Microsoft (before being reborn as the game-changing Firefox), using that to draw the absurd conclusion that if it were produced “properly” at the outset, today we’d all be using Netscape 9. Then again, maybe it would have followed the disastrous arc of Chandler.

The road that leads to most successful apps is often an ugly, brutish affair filled with compromise and folly, risk-taking, detours-followed, and shortcuts pursued. That isn’t to justify them, or to diminish alternative approaches, but we should always keep our minds open, being less quick to defensively guard whatever we’re selling as the cure-all this week.

A Call Out For Success Stories

What I’d like to read more about are the success stories, and less about the professional pundits telling everyone how it ought to be done.

Of course here’s where we get into a common paradox that exists in most industries: The successes are usually off enjoying their success and wealth, less inclined to toil away their days writing blog entries extolling their “dart toss” method of architecture. We’re left with the conversation being dominated by the people who
don’t actually make software at all, telling us how grand they could make software, if they ever actually did, by following their sure-win magic formula. The conversation boards are overrun with the people who actually have so little to do that they spend their time describing the ideal way that everyone else should write software.

Parallels can be drawn with the financial world, where the snake-oil salesmen pitching the ways to make money are usually doing so because their only way of making money is pitching how to make money (Want to know the secret to making big money Send me $5, and I’ll tell you that it’s to get people to send you $5 to learn the secret of making big money). The guys actually making the money are off making the money.

This brings us full circle back to Joel’s recommendation of the book. A book that serves as one of the few opportunities we have to really read how projects succeeded, straight from the source.

It’s good if only to let the successes have a voice in the conversation.