Paul Graham - Essays

Another way to burn up cycles is to have many layers of software between the application and the hardware. This too is a trend we see happening already: many recent languages are compiled into byte code. Bill Woods once told me that, as a rule of thumb, each layer of interpretation costs a factor of 10 in speed. This extra cost buys you flexibility.

The very first version of Arc was an extreme case of this sort of multi-level slowness, with corresponding benefits. It was a classic "metacircular" interpreter written on top of Common Lisp, with a definite family resemblance to the eval function defined in McCarthy's original Lisp paper. The whole thing was only a couple hundred lines of code, so it was very easy to understand and change. The Common Lisp we used, CLisp, itself runs on top of a byte code interpreter. So here we had two levels of interpretation, one of them (the top one) shockingly inefficient, and the language was usable. Barely usable, I admit, but usable.

Writing software as multiple layers is a powerful technique even within applications. Bottom-up programming means writing a program as a series of layers, each of which serves as a language for the one above. This approach tends to yield smaller, more flexible programs. It's also the best route to that holy grail, reusability. A language is by definition reusable. The more of your application you can push down into a language for writing that type of application, the more of your software will be reusable.

Somehow the idea of reusability got attached to object-oriented programming in the 1980s, and no amount of evidence to the contrary seems to be able to shake it free. But although some object-oriented software is reusable, what makes it reusable is its bottom-upness, not its object-orientedness. Consider libraries: they're reusable because they're language, whether they're written in an object-oriented style or not.

I don't predict the demise of object-oriented programming, by the way. Though I don't think it has much to offer good programmers, except in certain specialized domains, it is irresistible to large organizations. Object-oriented programming offers a sustainable way to write spaghetti code. It lets you accrete programs as a series of patches. Large organizations always tend to develop software this way, and I expect this to be as true in a hundred years as it is today.

As long as we're talking about the future, we had better talk about parallel computation, because that's where this idea seems to live. That is, no matter when you're talking, parallel computation seems to be something that is going to happen in the future.

Will the future ever catch up with it? People have been talking about parallel computation as something imminent for at least 20 years, and it hasn't affected programming practice much so far. Or hasn't it? Already chip designers have to think about it, and so must people trying to write systems software on multi-cpu computers.

The real question is, how far up the ladder of abstraction will parallelism go? In a hundred years will it affect even application programmers? Or will it be something that compiler writers think about, but which is usually invisible in the source code of applications?

One thing that does seem likely is that most opportunities for parallelism will be wasted. This is a special case of my more general prediction that most of the extra computer power we're given will go to waste. I expect that, as with the stupendous speed of the underlying hardware, parallelism will be something that is available if you ask for it explicitly, but ordinarily not used. This implies that the kind of parallelism we have in a hundred years will not, except in special applications, be massive parallelism. I expect for ordinary programmers it will be more like being able to fork off processes that all end up running in parallel.

And this will, like asking for specific implementations of data structures, be something that you do fairly late in the life of a program, when you try to optimize it. Version 1s will ordinarily ignore any advantages to be got from parallel computation, just as they will ignore advantages to be got from specific representations of data.

Except in special kinds of applications, parallelism won't pervade the programs that are written in a hundred years. It would be premature optimization if it did.

How many programming languages will there be in a hundred years? There seem to be a huge number of new programming languages lately. Part of the reason is that faster hardware has allowed programmers to make different tradeoffs between speed and convenience, depending on the application. If this is a real trend, the hardware we'll have in a hundred years should only increase it.

And yet there may be only a few widely-used languages in a hundred years. Part of the reason I say this is optimism: it seems that, if you did a really good job, you could make a language that was ideal for writing a slow version 1, and yet with the right optimization advice to the compiler, would also yield very fast code when necessary. So, since I'm optimistic, I'm going to predict that despite the huge gap they'll have between acceptable and maximal efficiency, programmers in a hundred years will have languages that can span most of it.

As this gap widens, profilers will become increasingly important. Little attention is paid to profiling now. Many people still seem to believe that the way to get fast applications is to write compilers that generate fast code. As the gap between acceptable and maximal performance widens, it will become increasingly clear that the way to get fast applications is to have a good guide from one to the other.

When I say there may only be a few languages, I'm not including domain-specific "little languages". I think such embedded languages are a great idea, and I expect them to proliferate. But I expect them to be written as thin enough skins that users can see the general-purpose language underneath.

Who will design the languages of the future? One of the most exciting trends in the last ten years has been the rise of open-source languages like Perl, Python, and Ruby. Language design is being taken over by hackers. The results so far are messy, but encouraging. There are some stunningly novel ideas in Perl, for example. Many are stunningly bad, but that's always true of ambitious efforts. At its current rate of mutation, God knows what Perl might evolve into in a hundred years.

It's not true that those who can't do, teach (some of the best hackers I know are professors), but it is true that there are a lot of things that those who teach can't do. Research imposes constraining caste restrictions. In any academic field there are topics that are ok to work on and others that aren't. Unfortunately the distinction between acceptable and forbidden topics is usually based on how intellectual the work sounds when described in research papers, rather than how important it is for getting good results. The extreme case is probably literature; people studying literature rarely say anything that would be of the slightest use to those producing it.

Though the situation is better in the sciences, the overlap between the kind of work you're allowed to do and the kind of work that yields good languages is distressingly small. (Olin Shivers has grumbled eloquently about this.) For example, types seem to be an inexhaustible source of research papers, despite the fact that static typing seems to preclude true macros-- without which, in my opinion, no language is worth using.

The trend is not merely toward languages being developed as open-source projects rather than "research", but toward languages being designed by the application programmers who need to use them, rather than by compiler writers. This seems a good trend and I expect it to continue.