Introduction-2

Richard Gabriel

Repetition, Generativity, and Patterns

Repetition---a Foucault pendulum, seawaves north of Santa Cruz, the shapely, curved stride of a young woman---attracts us---a steady R&B beat, night lullabies. Repetition---rhyming poetry, pre-dawn coffee---defines our lives---suppertime, weekends, evenings out.

When we were born, the world, fresh, contained no repetitions. An infant first smiles at his mother's return not because it is her but because her presence repeats; later he smiles because the she who returns is the same she. I take my daughter to the bus stop at 8:20 am every day, and as we stop she waits a minute then looks up the road to see the bus come over the ridgetop. She expects the bus at a particular time and feels safe and human when it does---out of sorts and stressed when it does not.

We understand all there is for us to understand through repeated parts and portions. Grammar is our regularized grunting, song is our stylized noisemaking, human society is our patterned gene stuff. We live by recognizing and using recurrences, by relying on what happens over and over. Unpredictability when not sought is feared.

Recurrence works because actions, characteristics, and relations that are noticeable repeat. Every act---every thing---contains not only the noticed repeated parts but also attendant, variable portions. The recurrent parts stand out because our minds (our brains?) are constructed that way.

Repetition as mantra is how we learn: 2x8 is 16, 3x8 is 24, 4x8 is 32. When I spelled ``stillness'' wrong in the 5th grade, I wrote it 100 times on the board. When I play squash and mis-hit or misdirect the ball, I chant, ``follow through, follow through, you idiot.'' A mantra of simple things. The magic of the chant---enchantment; inward song---incantation. When the rhythm of repetition begins, the body sways; when the repetition of chant continues, sometimes the chanter cannot stop. Repetition is power.

What repeats is important and reveals commonality. What isn't repeated is variable and can be highlighted or hidden. Repetition makes us feel secure and variation makes us feel free. Abstraction reveals and revels in what is common, the repetition, and what is abstract is defined by the repeated. The remainder, in science, basks behind the veil of ignorance, and the remainder, in computing, is filled in as missing. Abstractions is what we know as long as what is repeated has the magic of enchantment and song of incantation.

But, what of the mantra ``follow through, follow through''? Insistence of rhythm is a mnemonic---in ancient times laws were expressed in incantatory rhythms---and remembering and teaching are what we intend by the incantation. The law of follow-through is hard to forget---it's a law, though, that makes little sense: where is its enchantment? How can advice about what to do after you hit a ball make any difference to what happens when you hit it, or before?

The advice must work---because we hear it all the time---but how can it work? Here's how: If you plan to contact the ball at a particular point, your muscles will naturally begin slowing down just before you get there---perhaps to guard against the jolt. When you try to slow down the forward motion of your arm, muscles opposed to that motion contract to counteract. Your arm, then, moves with ``noise'', jiggling up and down, circling around; and power moving forward drops. Amplify this by the length of the racket. Problems.

When you follow through, your arm is aiming at a point well beyond where the racket will contact the ball, so there is no deceleration. ``Don't decelerate, don't decelerate, you idiot!'' It might make a mantra worth repeating, but it isn't advice you can take.

This mantra generates the effect we want---just as a seed generates the flower which eventually blossoms. Just as the wind in the sand generates dune designs and sidewinding sinews and ripples of grains of sand. The power of generative activities is cousin to that of repetition. Give a person something that can be done, and it will be done. Science loves a clever explanation---the gyroscopic forces generated by the front wheel of a bicycle cause the proper lean when you turn---but the clever explanation is merely clever, and really only good for a pleasant magazine article or MIT macho talk.

Generative mantras have the psychological power of ritual coupled with the physical power of algorithm. Ritual builds on our need for repetition and predictability, ritual is the basis of superstition---if you step with your left foot at the top of the dugout, rub toward the business end of the bat with the pine rag, dig a 3" trench to stand in with your right toe, you will not strike out. Ritual calms---the ritual of petting a dog lowers blood pressure below resting levels, makes stutterers speak clearly.

Rituals, repetitions, recurrences: These are simply patterns, the repeated parts, the familiar signposts that tame frightening variations. You don't know what the pitcher will throw, but you know how you will prepare as you walk to the plate. Rituals, like the advice to follow through, are things people can follow.

Software---is it something we can learn to produce by scientific study? Will type systems with nice semantics make it easy to produce and maintain the 5-million-line telecom switching system? Will an additional 30% in performance from unboxing data in a higher-order programming language unleash productivity in the softwareworkplace?

I don't think so. Such work will help a few dedicated researchers achieve tenure and assure a comfortable life for their long-enduring families, but you might as well tell a novice tennis player about how muscles decelerate and ask them to repeat their new stroke 20 times.

The work of the Hillside group is to remind us that people in teams and groups write software, and that in the absence of rituals and repetitions, recurrences, every step of the way for the individual is invention and creation. Software engineering is not yet engineering and won't be, cannot be, for decades if not centuries, because we cannot yet recognize the important, repeatable parts. When we do we will have patterns---recurrences, predictability.

Is it really possible we cannot yet recognize the repeatable parts? For several years I traveled to Paris every few months, and my good friend, Jerome, would always take me to the brasserie where he'd order oysters. Whenever I tasted them, all I could discern was salt water, fat, and sometimes grit. Jerome didn't skimp, and many times we had Locmariaquer oysters, but I ate the smallest polite number I could get away with.

Years later I visited Point Reyes Station in California, and my friend and I ordered oysters---Hog Island Sweetwaters. They're farmed in Tomales Bay where the river enters the bay, and they are only slightly salty. Suddenly I could taste oystertaste, sweet and lightly eccentric, muscularly hard, central. The next day---guarding against oddities---we ate oysters somewhere else, and once more I enjoyed them. The next year or so I sampled oysters whenever I could, and even when they weren't the Sweetwaters, I could taste the hint of oystertaste and could enjoy that part of it---the repeated part.

Two years ago Jerome and I were in Portland, Oregon, and I took him to Jake's where we dined on 5 dozen oysters of many types. He smiled and ordered champagne for the two of us.

I needed to learn to recognize the repeated parts---it took years and luck. Something as basic as flavor---I had to learn to taste it. Before that, oysters tasted like poison.

Software tastes like poison. Patterns, possibly, will reveal the sweetwaters.

Patterns embody the repetitions and recurrences within software; they provide the lonely developer with rituals and familiarity, and prevent him or her from falling prey to lossage. A pattern is a simple thing to understand, the more so the more like a ritual it makes part of development. Patterns feed on repetition, provide ritual. What patterns leave out provdes freedom, a point of departure for creativity.

There is no content to patterns---no one will ever produce a journal article or conference paper whose content is a pattern or pattern language---and if someone does, he's missed the point. Just as no one will ever write an acceptable biomechanics paper whose content is the advice to follow through.

If the ballplayer's ritual works it's because it relaxes him or gives him confidence. Patterns should be like these rituals---people should be able to follow them and perhaps there should rarely be reasons why or the reasons can be the topic of a real scientific paper. The content is not in the pattern---that's somewhere else---;the pattern holds only the ritual part, and the pattern works by giving the developer something to do with the knowledge only that ``if you do this, it will work.'' Such patterns are called generative patterns. Patterns work like rituals, patterns feed our need for repetition and predictability, patterns are designed for people to use in their lives---generative patterns work like magic. Generative patterns don't require CS degrees to understand, to get their gist you don't need a mathematical background---you just do what they say and win.

How do you produce generative patterns? You don't---you find them. They are in systems written by virtuosos, they are in systems that are a dream to maintain. If a great programmer comes to you and says ``hey, look at this code I found,'' that's where you'll be able to mine generative patterns. Someone invented or developed the code that contains the patterns, but that's largely irrelevant: It's that the same solution or style has been used and found habitable. We don't find a single gold atom, we find veins and nuggets---accumulation. Patterns are discovered because they are well-used, they form a vein that anyone could find.

Great pattern writers are miners, they create nothing except the wonderful explanation---they are writers, they aren't scientists, or even engineers. CS departments consider them drudges, scribes, amanuenses. Pattern writing geniuses won't get tenure anywhere, they won't advise CEO's. A pattern writer won't ride off into the sunset with the prize in his saddlebag

When we look at Christopher Alexander's patterns, we see that almost all of them have to do with how people live in homes, towns, and cities, with other people and alone, indulging in culture or savoring spirituality. Few talk about engineering and construction. They are short on physical explanation---we don't learn why light in a room causes useful biological reactions, only that such places are alive and comfortable. Alexander draws on centuries and millennia of architecture, and he, with a group of about a dozen, spent nearly a decade finding out 253 patterns.

We hunt today alone for software patterns in 40 years of code. Have we found any patterns yet?

Anyone can write lines in verse with a rhythm and some rhyme---is it poetry? Patterns are a form---the repetition calms, the variation inspires---but are we still writing doggerel?

Kent Beck

Once there was a mountain kingdom whose economy was based on farming. Wanting to get the most out of the limited arable land, the king set up schools to teach all the inhabitants the principles of farming---photosynthesis, weather, entymology. Upon graduation, each young lad or lass was assigned a patch of land to farm and given seeds and tools. Each was allowed, even encouraged, to farm their own way.

The results were disaster. Understanding the principles did not keep new farmers from making terrible mistakes. Crops were planted at the wrong time, crops in adjacent fields did not work together to limit the depredations of insects, the soil was not maintained from year to year.

As the food supply shrank, the country's residents became restless. They brought their empty bellies before the king and demanded a solution. He in turn brought together his wisest advisers and gave them a choice---solve the farming problem, or learn to do without their heads.

After many late night meetings where gallons of stimulating drinks were consumed, the advisers reported to the king. "King," they said, "the education given future farmers prepares them to farm in any number of ways. Toothless old farmers know, however, that only a few farming techniques deserve consideration. While understanding the principles is important, understanding the techniques that work is equally important. Bring together your most experienced and thoughtful farmers. Have them write down what works in a book. Give this book to new farmers, that they may avoid obvious mistakes and get on with farming."

At first the book grew quickly, as new farming techniques were recognized and recorded. It looked as if the book would grow so large that no one would be able to understand it all. After a while, though, it became apparent that many techniques were duplicates of others, and that the recognition of new techniques had slowed to a trickle. The book settled down to a weighty but manageable tome.

As farmers learned to read and use the book, the kingdom's food production stabilized. Farmers could now take advantage of the wisdom of others. When they faced a situation, they weren't responsible for always inventing a unique solution, they could find a technique that solved the problem. If no technique fit exactly, they were still able to adapt techniques rather than start from scratch.

The farmers sometimes missed the crazy old days of wild invention. On the whole, though, the kingdom was satisfied. The people ate well, so they were happy. The king didn't have to listen to complaints all day, so he was happy. And the advisers had solved the problem, so they still had heads to be happy with.

The Beginning