by Sid Holt
You keep your work in perspective by thinking about barrels of cash. You once heard that a U.S. dry barrel can hold about $100,000 worth of singles. Next year, you’ll burn a little under a barrel of cash on Oracle. One barrel isn’t that bad. But it’s never one barrel. Is this a five-barrel project or a ten-barreler? More? Too soon to tell. But you can definitely smell money burning.
At this stage in the meeting, you like to look supplicants in the eye and say, OK, you’ve given me a date and a budget. But when will it be done? Really, truly, top-line-revenue-reporting finished? Come to confession; unburden your soul.
This time you stop yourself. You don’t want your inquiry to be met by a patronizing sigh of impatience or another explanation about ship dates, Agile cycles, and continuous delivery. Better for now to hide your ignorance. When will it be done? You are learning to accept that the answer for software projects is never.
1.1. Why Are We Here?
We are here because the editor of this magazine asked me, “Can you tell me what code is?”
“No,” I said. “First of all, I’m not good at the math. I’m a programmer, yes, but I’m an East Coast programmer, not one of these serious platform people from the Bay Area.”
I began to program nearly twenty years ago, learning via oraperl, a special version of the Perl language modified to work with the Oracle database. A month into the work, I damaged the accounts of 30,000 fantasy basketball players. They sent some angry e-mails. After that, I decided to get better.
Which is to say I’m not a natural. I love computers, but they never made any sense to me. And yet, after two decades of jamming information into my code-resistant brain, I’ve amassed enough knowledge that the computer has revealed itself. Its magic has been stripped away. I can talk to someone who used to work at Amazon.com or Microsoft about his or her work without feeling a burning shame. I’d happily talk to people from Google and Apple, too, but they so rarely reenter the general population.
The World Wide Web is what I know best (I’ve coded for money in the programming languages Java, JavaScript, Python, Perl, PHP, Clojure, and XSLT), but the Web is only one small part of the larger world of software development. There are 11 million professional software developers on earth, according to the research firm IDC. (An additional 7 million are hobbyists.) That’s roughly the population of the greater Los Angeles metro area. Imagine all of L.A. programming. East Hollywood would be for Mac programmers, West L.A. for mobile, Beverly Hills for finance programmers, and all of Orange County for Windows.
There are lots of other neighborhoods, too: There are people who write code for embedded computers smaller than your thumb. There are people who write the code that runs your TV. There are programmers for everything. They have different cultures, different tribal folklores, that they use to organize their working life. If you told me a systems administrator was taking a juggling class, that would make sense, and I’d expect a product manager to take a trapeze class. I’ve met information architects who list and rank their friendships in spreadsheets. Security research specialists love to party.
What I’m saying is, I’m one of 18 million. So that’s what I’m writing: my view of software development, as an individual among millions. Code has been my life, and it has been your life, too. It is time to understand how it all works.
Every month it becomes easier to do things that have never been done before, to create new kinds of chaos and find new kinds of order. Even though my math skills will never catch up, I love the work. Every month, code changes the world in some interesting, wonderful, or disturbing way.
2. Let’s Begin
A computer is a clock with benefits. They all work the same, doing second-grade math, one step at a time: Tick, take a number and put it in box one. Tick, take another number, put it in box two. Tick, operate (an operation might be addition or subtraction) on those two numbers and put the resulting number in box one. Tick, check if the result is zero, and if it is, go to some other box and follow a new set of instructions.
You, using a pen and paper, can do anything a computer can; you just can’t do those things billions of times per second. And those billions of tiny operations add up. They can cause a phone to boop, elevate an elevator, or redirect a missile. That raw speed makes it possible to pull off not one but multiple sleights of hand, card tricks on top of card tricks. Take a bunch of pulses of light reflected from an optical disc, apply some math to unsqueeze them, and copy the resulting pile of expanded impulses into some memory cells—then read from those cells to paint light on the screen. Millions of pulses, sixty times a second. That’s how you make the rubes believe they’re watching a movie.
Apple has always made computers; Microsoft used to make only software (and occasional accessory hardware, such as mice and keyboards), but now it’s in the hardware business, with Xbox game consoles, Surface tablets, and Lumia phones. Facebook assembles its own computers for its massive data centers.
So many things are computers, or will be. That includes watches, cameras, air conditioners, cash registers, toilets, toys, airplanes, and movie projectors. Samsung makes computers that look like TVs, and Tesla makes computers with wheels and engines. Some things that aren’t yet computers—dental floss, flashlights—will fall eventually.
When you “batch” process a thousand images in Photoshop or sum numbers in Excel, you’re programming, at least a little. When you use computers too much—which is to say a typical amount—they start to change you. I’ve had Photoshop dreams, Visio dreams, spreadsheet dreams, and web-browser dreams. The dreamscape becomes fluid and can be sorted and restructured. I’ve had programming dreams where I move text around the screen.
You can make computers do wonderful things, but you need to understand their limits. They’re not all-powerful, not conscious in the least. They’re fast, but some parts—the processor, the RAM—are faster than others—like the hard drive or the network connection. Making them seem infinite takes a great deal of work from a lot of programmers and a lot of marketers.
The turn-of-last-century British artist William Morris once said you can’t have art without resistance in the materials. The computer and its multifarious peripherals are the materials. The code is the art.
2.1. How Do You Type an “A”?
Consider what happens when you strike a key on your keyboard. Say a lowercase “a.” The keyboard is waiting for you to press a key or release one; it’s constantly scanning to see what keys are pressed down. Hitting the key sends a scancode.
Just as the keyboard is waiting for a key to be pressed, the computer is waiting for a signal from the keyboard. When one comes down the pike, the computer interprets it and passes it farther into its own interior. “Here’s what the keyboard just received—do with this what you will.”
It’s simple now, right? The computer just goes to some table, figures out that the signal corresponds to the letter “a,” and puts it on screen. Of course not—too easy. Computers are machines. They don’t know what a screen or an “a” are. To put the “a” on the screen, your computer has to pull the image of the “a” out of its memory as part of a font, an “a” made up of lines and circles. It has to take these lines and circles and render them in a little box of pixels in the part of its memory that manages the screen. So far we have at least three representations of one letter: the signal from the keyboard; the version in memory; and the lines-and-circles version sketched on the screen. We haven’t even considered how to store it or what happens to the letters to the left and the right when you insert an “a” in the middle of a sentence. Or what “lines and circles” mean when reduced to binary data. There are surprisingly many ways to represent a simple “a.” It’s amazing any of it works at all.
Coders are people who are willing to work backward to that key press. It takes a certain temperament to page through standards documents, manuals, and documentation and read things like “data fields are transmitted least significant bit first” in the interest of understandi
ng why, when you expected “ü,” you keep getting “.”
2.2. From Hardware to Software
Hardware is a tricky business. For decades the work of integrating, building, and shipping computers was a way to build fortunes. But margins tightened. Look at Dell, now back in private hands, or Gateway, acquired by Acer. Dell and Gateway, two world-beating companies, stayed out of software, typically building PCs that came preinstalled with Microsoft Windows—plus various subscription-based services to increase profits.
This led to much cursing from individuals who’d spent $1,000 or more on a computer and now had to figure out how to stop the antivirus software from nagging them to pay up.
Years ago, when Microsoft was king, Steve Ballmer, sweating through his blue button-down, jumped up and down in front of a stadium full of people and chanted, “Developers! Developers! Developers! Developers!”
He yelled until he was hoarse: “I love this company!” Of course he did. If you can sell the software, if you can light up the screen, you’re selling infinitely reproducible nothings. The margins on nothing are great—until other people start selling even cheaper nothings or giving them away. Which is what happened, as free software-based systems such as Linux began to nibble, then devour, the server market, and free-to-use web-based applications such as Google Apps began to serve as viable replacements for desktop software.
Expectations around software have changed over time. IBM unbundled software from hardware in the 1960s and got to charge more; Microsoft rebundled Internet Explorer with Windows in 1998 and got sued; Apple initially refused anyone else the ability to write software for the iPhone when it came out in 2007, and then opened the App Store, which expanded into a vast commercial territory—and soon the world had Angry Birds. Today, much hardware comes with some software—a PC comes with an operating system, for example, and that OS includes hundreds of subprograms, from mail apps to solitaire. Then you download or buy more.
There have been countless attempts to make software easier to write, promising that you could code in plain English, or manipulate a set of icons, or make a list of rules—software development so simple that a bright senior executive or an average child could do it. Decades of efforts have gone into helping civilians write code as they might use a calculator or write an e-mail. Nothing yet has done away with developers, developers, developers, developers.
Thus a craft, and a professional class that lives that craft, emerged. Beginning in the 1950s, but catching fire in the 1980s, a proportionally small number of people became adept at inventing ways to satisfy basic human desires (know the time, schedule a flight, send a letter, kill a zombie) by controlling the machine. Coders, starting with concepts such as “signals from a keyboard” and “numbers in memory,” created infinitely reproducible units of digital execution that we call software, hoping to meet the needs of the marketplace. Man, did they. The systems they built are used to manage the global economic infrastructure. If coders don’t run the world, they run the things that run the world.
Most programmers aren’t working on building a widely recognized application like Microsoft Word. Software is everywhere. It’s gone from a craft of fragile, built-from-scratch custom projects to an industry of standardized parts, where coders absorb and improve upon the labors of their forebears (even if those forebears are one cubicle over). Software is there when you switch channels and your cable box shows you what else is on. You get money from an ATM—software. An elevator takes you up five stories—the same. Facebook releases software every day to something like a billion people, and that software runs inside web browsers and mobile applications. Facebook looks like it’s just pictures of your mom’s crocuses or your son’s school play—but no, it’s software.
2.3. How Does Code Become Software?
We know that a computer is a clock with benefits, and that software starts as code, but how?
We know that someone, somehow, enters a program into the computer and the program is made of code. In the old days, that meant putting holes in punch cards. Then you’d put the cards into a box and give them to an operator who would load them, and the computer would flip through the cards, identify where the holes were, and update parts of its memory, and then it would—OK, that’s a little too far back. Let’s talk about modern typing-into-a-keyboard code. It might look like this:
ispal: {x~|x}
That’s in a language called, simply, K, famous for its brevity. That code will test if something is a palindrome. If you next typed in ispal “able was i ere i saw elba”, K will confirm that yes, this is a palindrome.
Coding is a broad human activity, like sport or writing. When software developers think of coding, most of them are thinking about lines of code in files. They’re handed a problem, think about the problem, write code that will solve the problem, and then expect the computer to turn word into deed.
Code is inert. How do you make it ert? You run software that transforms it into machine language. The word “language” is a little ambitious here, given that you can make a computing device with wood and marbles. Your goal is to turn your code into an explicit list of instructions that can be carried out by interconnected logic gates, thus turning your code into something that can be executed—software.
Computing treats human language as an arbitrary set of symbols in sequences. It treats music, imagery, and film that way, too. It’s a good and healthy exercise to ponder what your computer is doing right now. Maybe you’re reading this on a laptop: What are the steps and layers between what you’re doing and the Lilliputian mechanisms within? When you double-click an icon to open a program such as a word processor, the computer must know where that program is on the disk. It has some sort of accounting process to do that. And then it loads that program into its memory—which means that it loads an enormous to-do list into its memory and starts to step through it. What does that list look like?
Maybe you’re reading this in print. No shame in that. In fact, thank you. The paper is the artifact of digital processes. Remember how we put that “a” on screen? See if you can get from some sleepy writer typing that letter on a keyboard in Brooklyn, N.Y., to the paper under your thumb. What framed that fearful symmetry?
Thinking this way will teach you two things about computers: One, there’s no magic, no matter how much it looks like there is. There’s just work to make things look like magic. And two, it’s crazy in there.
2.4. What Is an Algorithm?
“Algorithm” is a word writers invoke to sound smart about technology. Journalists tend to talk about “Facebook’s algorithm” or a “Google algorithm,” which is usually inaccurate. They mean “software.”
Algorithms don’t require computers any more than geometry does. An algorithm solves a problem, and a great algorithm gets a name. Euclid’s algorithm, for example, is the go-to specimen that shows up whenever anyone wants to wax on about algorithms, so why buck the trend? It’s a simple way of determining the greatest common divisor for two numbers. Take two numbers, like 16 and 12. Divide the first by the second. If there’s a remainder (in this case there is, 4), divide the smaller number, 12, by that remainder, 4, which gives you 3 and no remainder, so we’re done—and 4 is the greatest common divisor. (Now translate that into machine code, and we can get out of here.)
A programming language is a system for encoding, naming, and organizing algorithms for reuse and application. It’s an algorithm-management system. This is why, despite the hype, it’s silly to say Facebook has an algorithm. An algorithm can be translated into a function, and that function can be called (run) when software is executed. There are algorithms that relate to image processing and for storing data efficiently and for rapidly running through the elements of a list. Most algorithms come for free, already built into a programming language, or are available, organized into libraries, for download from the Internet in a moment. You can do a ton of programming without actually thinking about algorithms—you can save something into a database or print a webpag
e by cutting and pasting code. But if you want the computer to, say, identify whether it’s reading Spanish or Italian, you’ll need to write a language-matching function. So in that sense, algorithms can be pure, mathematical entities as well as practical expressions of ideas on which you can place your grubby hands.
One thing that took me forever to understand is that computers aren’t actually “good at math.” They can be programmed to execute certain operations to certain degrees of precision, so much so that it looks like “doing math” to humans. A huge part of computer science is about understanding the efficiency of algorithms—how long they will take to run. Computers are fast, but they can get bogged down—for example, when trying to find the shortest path between two points on a large map. Companies such as Google, Facebook, and Twitter are built on top of fundamental computer science and pay great attention to efficiency because their users do things (searches, status updates, tweets) an extraordinary number of times. Thus it’s absolutely worth their time to find excellent computer scientists, many with doctorates, who know where all the efficiencies are buried.
It takes a good mathematician to be a computer scientist, but a middling one to be an effective programmer. Until you start dealing with millions of people on a network or you need to blur or sharpen a million photos quickly, you can just use the work of other people. When it gets real, break out the comp sci. When you’re doing anything a hundred trillion times, nanosecond delays add up. Systems slow down, users get cranky, money burns by the barrel.
The hardest work in programming is getting around things that aren’t computable, in finding ways to break impossible tasks into small, possible components, and then creating the impression that the computer is doing something it actually isn’t, like having a human conversation. This used to be known as “artificial intelligence research,” but now it’s more likely to go under the name “machine learning” or “data mining.” When you speak to Siri or Cortana and they respond, it’s not because these services understand you; they convert your words into text, break that text into symbols, then match those symbols against the symbols in their database of terms, and produce an answer. Tons of algorithms, bundled up and applied, mean that computers can fake listening.
