☰

5.5 & 5.6 Liquid infrastructure

“Enterprise” is a feared word among programmers, because enterprise programming is a lot of work without much to show for it. Remember healthcare.gov, the first version that was a total disaster? Perfect example of enterprise coding. At the same time, programmers respect big systems—when they work. We respect the ambition of huge heavy machines running big blobs of code. We grew up reading about supercomputers. Big iron is cool, even if the future seems to be huge cloud platforms hosting with tons of cheap computers.

But Java is also in wide use at Google. It’s a language for places such as General Electric and Accenture. These aren’t startups, but if their product schedules slip, so does their revenue, and they are beholden to the public markets. Gigantic data-driven organizations are structured around code, around getting software made. But that doesn’t mean their teams are huge—Amazon, for example, is famous for its two-pizza rule: “Never have a meeting where two pizzas couldn’t feed the entire group.”

These companies have cultures that know how to make software. They have whole departments dedicated to testing. The process is important because there are so many moving pieces, many of them invisible.

Academic researchers often produce things that basically work but don’t have interfaces. They need to prove their theses, publish, and move on to the next thing. People in the free software community often code to scratch an itch and release that code into the digital commons so that other people can modify and manipulate it. While more often than not this process goes nowhere, over time some projects capture the imagination of others and become part of the infrastructure of the world.

Java, interestingly, profits from all this. It’s designed for big corporate projects and has the infrastructure to support them. It’s also a useful language for midsize tasks. So the libraries that you need to do things—image processing, logging on to files, full-text search—keep appearing at a steady clip, improving on the standard libraries or supplanting them entirely.

Eventually, people realized that if they didn’t like the Java language, they could write other languages that compile to Java “bytecode” and run on the Java virtual machine (JVM). So there are now many languages that run on top of Java. Some are versions of well-known languages, such as Jython and JRuby. Others are totally new, like Scala, which is one of the languages that Twitter began to use when it outgrew Ruby. There’s also Clojure, which is … a Lisp. That is, Clojure takes a much-adored computer language that was born in the 1950s, updates it for the 2010s, and gives that language the ability to reuse all the Java classes in the world.

Clojure was created by Rich Hickey, who decided that he wanted a new, modern version of Lisp.26 Lisp may be old for a language, but it’s still revered as a classic model for thought, like a Braun shaver or an Eames chair for the mind. Hickey worked on Clojure for many years, eventually adapting it to run on the JVM. This was a dramatic choice, because it meant that Clojure had access to the entirety of the Java class library.

Now you walk into the office and sit at your computer and write a little Lisp-ish code, very elegant and well-considered and trim and comfortable, but you have access to thousands and thousands of incredibly well-thought-out and totally functional free libraries that you can use to get your work done.

Anything Java can do, Clojure can do. And since it’s built atop the JVM, it can do it on any computer. There were already Lisp editing tools out there, and it was pretty easy to modify them for Clojure. It was joined to Java like a remora to a shark. Or more accurately, it’s a remora attached to a remora, because the JVM itself is a fake machine running inside real machines.

Clojure is beloved but not widely used. Perhaps it will become more popular. It shows up in interesting places. It’s used at Wal-Mart Stores to collect and manage receipt data, for example. It was ported to the world of Microsoft and runs right inside Excel. It’s the craziest thing. Take a spreadsheet. Inside the spreadsheet is some Clojure code just running away, addressing the different cells and columns. Someone made a version that compiles (well, transpiles) to JavaScript, so blam, Clojure’s running in your Web browser. Actually that version is called ClojureScript.

The point is that things are fluid in the world of programming, fluid in a way that other industries don’t seem to be. Languages are liquid infrastructure. You download a few programs and, whoa, suddenly you have a working Clojure environment. Which is actually the Java Runtime Environment. You grab an old PC that’s outlived its usefulness, put Linux on it, and suddenly you have a powerful Web server. Now you can participate in whole new cultures. There are meetups, gatherings, conferences, blogs, and people chatting on Twitter. And you are welcomed. They are glad for the new blood.

Java was supposed to supplant C and run on smart jewelry. Now it runs application servers, hosts Lisplike languages, and is the core language of the Android operating system.27 It runs on billions of things. It won. C and C++, which it was designed to supplant, also won. A lot of things keep winning because computers keep getting more plentiful. It’s weird.

Off the Shelf

A few weeks later, when he’s in the office and in for his meeting, you ask TMitTB if there’s a way to use more off-the-shelf components, a way to buy your way out of this.

He makes a face. Sort of, he says, but when you’re making a system that will integrate with the systems around it and your company is a set of such systems, nothing is truly off the shelf. There are tools and packages and libraries, and if you have any wit at all you already use well-documented, free code for things such as e-mail validation, but that obviates only so much.

“Everything is edge cases,” he says. “Testing and edge cases.”