# RSS

[](index.html)  
  

  
  
Aaron Swartz created a scraped [feed](http://www.aaronsw.com/2002/feeds/pgessays.rss) of the essays page.  
  
  
  

  
  

* * *



# This Year We Can End the Death Penalty in California

[](index.html)  
  

  
  
November 2016  
  
If you're a California voter, there is an important proposition on your ballot this year: Proposition 62, which bans the death penalty.  
  
When I was younger I used to think the debate about the death penalty was about when it's ok to take a human life. Is it ok to kill a killer?  
  
But that is not the issue here.  
  
The real world does not work like the version I was shown on TV growing up. The police often arrest the wrong person. Defendants' lawyers are often incompetent. And prosecutors are often motivated more by publicity than justice.  
  
In the real world, [about 4%](http://time.com/79572/more-innocent-people-on-death-row-than-estimated-study/) of people sentenced to death are innocent. So this is not about whether it's ok to kill killers. This is about whether it's ok to kill innocent people.  
  
A child could answer that one for you.  
  
This year, in California, you have a chance to end this, by voting yes on Proposition 62. But beware, because there is another proposition, Proposition 66, whose goal is to make it easier to execute people. So yes on 62, no on 66.  
  
It's time.  
  

  
  

* * *



# Programming Bottom-Up

[](index.html)  
  

  
  
1993

_(This essay is from the introduction to_ [On Lisp](onlisp.html)_.)_  
  
It's a long-standing principle of programming style that the functional elements of a program should not be too large. If some component of a program grows beyond the stage where it's readily comprehensible, it becomes a mass of complexity which conceals errors as easily as a big city conceals fugitives. Such software will be hard to read, hard to test, and hard to debug.  
  
In accordance with this principle, a large program must be divided into pieces, and the larger the program, the more it must be divided. How do you divide a program? The traditional approach is called _top-down design:_ you say "the purpose of the program is to do these seven things, so I divide it into seven major subroutines. The first subroutine has to do these four things, so it in turn will have four of its own subroutines," and so on. This process continues until the whole program has the right level of granularity-- each part large enough to do something substantial, but small enough to be understood as a single unit.  
  
Experienced Lisp programmers divide up their programs differently. As well as top-down design, they follow a principle which could be called _bottom-up design_\-- changing the language to suit the problem. In Lisp, you don't just write your program down toward the language, you also build the language up toward your program. As you're writing a program you may think "I wish Lisp had such-and-such an operator." So you go and write it. Afterward you realize that using the new operator would simplify the design of another part of the program, and so on. Language and program evolve together. Like the border between two warring states, the boundary between language and program is drawn and redrawn, until eventually it comes to rest along the mountains and rivers, the natural frontiers of your problem. In the end your program will look as if the language had been designed for it. And when language and program fit one another well, you end up with code which is clear, small, and efficient.  
  
It's worth emphasizing that bottom-up design doesn't mean just writing the same program in a different order. When you work bottom-up, you usually end up with a different program. Instead of a single, monolithic program, you will get a larger language with more abstract operators, and a smaller program written in it. Instead of a lintel, you'll get an arch.  
  
In typical code, once you abstract out the parts which are merely bookkeeping, what's left is much shorter; the higher you build up the language, the less distance you will have to travel from the top down to it. This brings several advantages:  
  

1.  By making the language do more of the work, bottom-up design yields programs which are smaller and more agile. A shorter program doesn't have to be divided into so many components, and fewer components means programs which are easier to read or modify. Fewer components also means fewer connections between components, and thus less chance for errors there. As industrial designers strive to reduce the number of moving parts in a machine, experienced Lisp programmers use bottom-up design to reduce the size and complexity of their programs.  
      
    
2.  Bottom-up design promotes code re-use. When you write two or more programs, many of the utilities you wrote for the first program will also be useful in the succeeding ones. Once you've acquired a large substrate of utilities, writing a new program can take only a fraction of the effort it would require if you had to start with raw Lisp.  
      
    
3.  Bottom-up design makes programs easier to read. An instance of this type of abstraction asks the reader to understand a general-purpose operator; an instance of functional abstraction asks the reader to understand a special-purpose subroutine. \[1\]  
      
    
4.  Because it causes you always to be on the lookout for patterns in your code, working bottom-up helps to clarify your ideas about the design of your program. If two distant components of a program are similar in form, you'll be led to notice the similarity and perhaps to redesign the program in a simpler way.

Bottom-up design is possible to a certain degree in languages other than Lisp. Whenever you see library functions, bottom-up design is happening. However, Lisp gives you much broader powers in this department, and augmenting the language plays a proportionately larger role in Lisp style-- so much so that Lisp is not just a different language, but a whole different way of programming.  
  
It's true that this style of development is better suited to programs which can be written by small groups. However, at the same time, it extends the limits of what can be done by a small group. In _The Mythical Man-Month_, Frederick Brooks proposed that the productivity of a group of programmers does not grow linearly with its size. As the size of the group increases, the productivity of individual programmers goes down. The experience of Lisp programming suggests a more cheerful way to phrase this law: as the size of the group decreases, the productivity of individual programmers goes up. A small group wins, relatively speaking, simply because it's smaller. When a small group also takes advantage of the techniques that Lisp makes possible, it can [win outright](avg.html).  
  
  
  
**New:** [Download On Lisp for Free](onlisptext.html).  
  
  
  

* * *

  
  
\[1\] "But no one can read the program without understanding all your new utilities." To see why such statements are usually mistaken, see Section 4.8.  
  

  
  

* * *



# Lisp for Web-Based Applications

[](index.html)  
  

  
  

After a link to [Beating the Averages](avg.html) was posted on slashdot, some readers wanted to hear in more detail about the specific technical advantages we got from using Lisp in Viaweb. For those who are interested, here are some excerpts from a talk I gave in April 2001 at BBN Labs in Cambridge, MA.  
  

  

[BBN Talk Excerpts (ASCII)](https://sep.yimg.com/ty/cdn/paulgraham/bbnexcerpts.txt?t=1595850613&)  
  

  

  
  

* * *



# Beating the Averages

[](index.html)  
  

  
  

**Want to start a startup?** Get funded by [Y Combinator](http://ycombinator.com/apply.html).  

April 2001, rev. April 2003  
  
_(This article is derived from a talk given at the 2001 Franz Developer Symposium.)_  
  
In the summer of 1995, my friend Robert Morris and I started a startup called [Viaweb](http://docs.yahoo.com/docs/pr/release184.html). Our plan was to write software that would let end users build online stores. What was novel about this software, at the time, was that it ran on our server, using ordinary Web pages as the interface.  
  
A lot of people could have been having this idea at the same time, of course, but as far as I know, Viaweb was the first Web-based application. It seemed such a novel idea to us that we named the company after it: Viaweb, because our software worked via the Web, instead of running on your desktop computer.  
  
Another unusual thing about this software was that it was written primarily in a programming language called Lisp. It was one of the first big end-user applications to be written in Lisp, which up till then had been used mostly in universities and research labs. \[1\]  
  
**The Secret Weapon**  
  
Eric Raymond has written an essay called "How to Become a Hacker," and in it, among other things, he tells would-be hackers what languages they should learn. He suggests starting with Python and Java, because they are easy to learn. The serious hacker will also want to learn C, in order to hack Unix, and Perl for system administration and cgi scripts. Finally, the truly serious hacker should consider learning Lisp:

> Lisp is worth learning for the profound enlightenment experience you will have when you finally get it; that experience will make you a better programmer for the rest of your days, even if you never actually use Lisp itself a lot.

This is the same argument you tend to hear for learning Latin. It won't get you a job, except perhaps as a classics professor, but it will improve your mind, and make you a better writer in languages you do want to use, like English.  
  
But wait a minute. This metaphor doesn't stretch that far. The reason Latin won't get you a job is that no one speaks it. If you write in Latin, no one can understand you. But Lisp is a computer language, and computers speak whatever language you, the programmer, tell them to.  
  
So if Lisp makes you a better programmer, like he says, why wouldn't you want to use it? If a painter were offered a brush that would make him a better painter, it seems to me that he would want to use it in all his paintings, wouldn't he? I'm not trying to make fun of Eric Raymond here. On the whole, his advice is good. What he says about Lisp is pretty much the conventional wisdom. But there is a contradiction in the conventional wisdom: Lisp will make you a better programmer, and yet you won't use it.  
  
Why not? Programming languages are just tools, after all. If Lisp really does yield better programs, you should use it. And if it doesn't, then who needs it?  
  
This is not just a theoretical question. Software is a very competitive business, prone to natural monopolies. A company that gets software written faster and better will, all other things being equal, put its competitors out of business. And when you're starting a startup, you feel this very keenly. Startups tend to be an all or nothing proposition. You either get rich, or you get nothing. In a startup, if you bet on the wrong technology, your competitors will crush you.  
  
Robert and I both knew Lisp well, and we couldn't see any reason not to trust our instincts and go with Lisp. We knew that everyone else was writing their software in C++ or Perl. But we also knew that that didn't mean anything. If you chose technology that way, you'd be running Windows. When you choose technology, you have to ignore what other people are doing, and consider only what will work the best.  
  
This is especially true in a startup. In a big company, you can do what all the other big companies are doing. But a startup can't do what all the other startups do. I don't think a lot of people realize this, even in startups.  
  
The average big company grows at about ten percent a year. So if you're running a big company and you do everything the way the average big company does it, you can expect to do as well as the average big company-- that is, to grow about ten percent a year.  
  
The same thing will happen if you're running a startup, of course. If you do everything the way the average startup does it, you should expect average performance. The problem here is, average performance means that you'll go out of business. The survival rate for startups is way less than fifty percent. So if you're running a startup, you had better be doing something odd. If not, you're in trouble.  
  
Back in 1995, we knew something that I don't think our competitors understood, and few understand even now: when you're writing software that only has to run on your own servers, you can use any language you want. When you're writing desktop software, there's a strong bias toward writing applications in the same language as the operating system. Ten years ago, writing applications meant writing applications in C. But with Web-based software, especially when you have the source code of both the language and the operating system, you can use whatever language you want.  
  
This new freedom is a double-edged sword, however. Now that you can use any language, you have to think about which one to use. Companies that try to pretend nothing has changed risk finding that their competitors do not.  
  
If you can use any language, which do you use? We chose Lisp. For one thing, it was obvious that rapid development would be important in this market. We were all starting from scratch, so a company that could get new features done before its competitors would have a big advantage. We knew Lisp was a really good language for writing software quickly, and server-based applications magnify the effect of rapid development, because you can release software the minute it's done.  
  
If other companies didn't want to use Lisp, so much the better. It might give us a technological edge, and we needed all the help we could get. When we started Viaweb, we had no experience in business. We didn't know anything about marketing, or hiring people, or raising money, or getting customers. Neither of us had ever even had what you would call a real job. The only thing we were good at was writing software. We hoped that would save us. Any advantage we could get in the software department, we would take.  
  
So you could say that using Lisp was an experiment. Our hypothesis was that if we wrote our software in Lisp, we'd be able to get features done faster than our competitors, and also to do things in our software that they couldn't do. And because Lisp was so high-level, we wouldn't need a big development team, so our costs would be lower. If this were so, we could offer a better product for less money, and still make a profit. We would end up getting all the users, and our competitors would get none, and eventually go out of business. That was what we hoped would happen, anyway.  
  
What were the results of this experiment? Somewhat surprisingly, it worked. We eventually had many competitors, on the order of twenty to thirty of them, but none of their software could compete with ours. We had a wysiwyg online store builder that ran on the server and yet felt like a desktop application. Our competitors had cgi scripts. And we were always far ahead of them in features. Sometimes, in desperation, competitors would try to introduce features that we didn't have. But with Lisp our development cycle was so fast that we could sometimes duplicate a new feature within a day or two of a competitor announcing it in a press release. By the time journalists covering the press release got round to calling us, we would have the new feature too.  
  
It must have seemed to our competitors that we had some kind of secret weapon-- that we were decoding their Enigma traffic or something. In fact we did have a secret weapon, but it was simpler than they realized. No one was leaking news of their features to us. We were just able to develop software faster than anyone thought possible.  
  
When I was about nine I happened to get hold of a copy of _The Day of the Jackal,_ by Frederick Forsyth. The main character is an assassin who is hired to kill the president of France. The assassin has to get past the police to get up to an apartment that overlooks the president's route. He walks right by them, dressed up as an old man on crutches, and they never suspect him.  
  
Our secret weapon was similar. We wrote our software in a weird AI language, with a bizarre syntax full of parentheses. For years it had annoyed me to hear Lisp described that way. But now it worked to our advantage. In business, there is nothing more valuable than a technical advantage your competitors don't understand. In business, as in war, surprise is worth as much as force.  
  
And so, I'm a little embarrassed to say, I never said anything publicly about Lisp while we were working on Viaweb. We never mentioned it to the press, and if you searched for Lisp on our Web site, all you'd find were the titles of two books in my bio. This was no accident. A startup should give its competitors as little information as possible. If they didn't know what language our software was written in, or didn't care, I wanted to keep it that way.\[2\]  
  
The people who understood our technology best were the customers. They didn't care what language Viaweb was written in either, but they noticed that it worked really well. It let them build great looking online stores literally in minutes. And so, by word of mouth mostly, we got more and more users. By the end of 1996 we had about 70 stores online. At the end of 1997 we had 500. Six months later, when Yahoo bought us, we had 1070 users. Today, as Yahoo Store, this software continues to dominate its market. It's one of the more profitable pieces of Yahoo, and the stores built with it are the foundation of Yahoo Shopping. I left Yahoo in 1999, so I don't know exactly how many users they have now, but the last I heard there were about 20,000.  
  
**The Blub Paradox**  
  
What's so great about Lisp? And if Lisp is so great, why doesn't everyone use it? These sound like rhetorical questions, but actually they have straightforward answers. Lisp is so great not because of some magic quality visible only to devotees, but because it is simply the most powerful language available. And the reason everyone doesn't use it is that programming languages are not merely technologies, but habits of mind as well, and nothing changes slower. Of course, both these answers need explaining.  
  
I'll begin with a shockingly controversial statement: programming languages vary in power.  
  
Few would dispute, at least, that high level languages are more powerful than machine language. Most programmers today would agree that you do not, ordinarily, want to program in machine language. Instead, you should program in a high-level language, and have a compiler translate it into machine language for you. This idea is even built into the hardware now: since the 1980s, instruction sets have been designed for compilers rather than human programmers.  
  
Everyone knows it's a mistake to write your whole program by hand in machine language. What's less often understood is that there is a more general principle here: that if you have a choice of several languages, it is, all other things being equal, a mistake to program in anything but the most powerful one. \[3\]  
  
There are many exceptions to this rule. If you're writing a program that has to work very closely with a program written in a certain language, it might be a good idea to write the new program in the same language. If you're writing a program that only has to do something very simple, like number crunching or bit manipulation, you may as well use a less abstract language, especially since it may be slightly faster. And if you're writing a short, throwaway program, you may be better off just using whatever language has the best library functions for the task. But in general, for application software, you want to be using the most powerful (reasonably efficient) language you can get, and using anything else is a mistake, of exactly the same kind, though possibly in a lesser degree, as programming in machine language.  
  
You can see that machine language is very low level. But, at least as a kind of social convention, high-level languages are often all treated as equivalent. They're not. Technically the term "high-level language" doesn't mean anything very definite. There's no dividing line with machine languages on one side and all the high-level languages on the other. Languages fall along a continuum \[4\] of abstractness, from the most powerful all the way down to machine languages, which themselves vary in power.  
  
Consider Cobol. Cobol is a high-level language, in the sense that it gets compiled into machine language. Would anyone seriously argue that Cobol is equivalent in power to, say, Python? It's probably closer to machine language than Python.  
  
Or how about Perl 4? Between Perl 4 and Perl 5, lexical closures got added to the language. Most Perl hackers would agree that Perl 5 is more powerful than Perl 4. But once you've admitted that, you've admitted that one high level language can be more powerful than another. And it follows inexorably that, except in special cases, you ought to use the most powerful you can get.  
  
This idea is rarely followed to its conclusion, though. After a certain age, programmers rarely switch languages voluntarily. Whatever language people happen to be used to, they tend to consider just good enough.  
  
Programmers get very attached to their favorite languages, and I don't want to hurt anyone's feelings, so to explain this point I'm going to use a hypothetical language called Blub. Blub falls right in the middle of the abstractness continuum. It is not the most powerful language, but it is more powerful than Cobol or machine language.  
  
And in fact, our hypothetical Blub programmer wouldn't use either of them. Of course he wouldn't program in machine language. That's what compilers are for. And as for Cobol, he doesn't know how anyone can get anything done with it. It doesn't even have x (Blub feature of your choice).  
  
As long as our hypothetical Blub programmer is looking down the power continuum, he knows he's looking down. Languages less powerful than Blub are obviously less powerful, because they're missing some feature he's used to. But when our hypothetical Blub programmer looks in the other direction, up the power continuum, he doesn't realize he's looking up. What he sees are merely weird languages. He probably considers them about equivalent in power to Blub, but with all this other hairy stuff thrown in as well. Blub is good enough for him, because he thinks in Blub.  
  
When we switch to the point of view of a programmer using any of the languages higher up the power continuum, however, we find that he in turn looks down upon Blub. How can you get anything done in Blub? It doesn't even have y.  
  
By induction, the only programmers in a position to see all the differences in power between the various languages are those who understand the most powerful one. (This is probably what Eric Raymond meant about Lisp making you a better programmer.) You can't trust the opinions of the others, because of the Blub paradox: they're satisfied with whatever language they happen to use, because it dictates the way they think about programs.  
  
I know this from my own experience, as a high school kid writing programs in Basic. That language didn't even support recursion. It's hard to imagine writing programs without using recursion, but I didn't miss it at the time. I thought in Basic. And I was a whiz at it. Master of all I surveyed.  
  
The five languages that Eric Raymond recommends to hackers fall at various points on the power continuum. Where they fall relative to one another is a sensitive topic. What I will say is that I think Lisp is at the top. And to support this claim I'll tell you about one of the things I find missing when I look at the other four languages. How can you get anything done in them, I think, without macros? \[5\]  
  
Many languages have something called a macro. But Lisp macros are unique. And believe it or not, what they do is related to the parentheses. The designers of Lisp didn't put all those parentheses in the language just to be different. To the Blub programmer, Lisp code looks weird. But those parentheses are there for a reason. They are the outward evidence of a fundamental difference between Lisp and other languages.  
  
Lisp code is made out of Lisp data objects. And not in the trivial sense that the source files contain characters, and strings are one of the data types supported by the language. Lisp code, after it's read by the parser, is made of data structures that you can traverse.  
  
If you understand how compilers work, what's really going on is not so much that Lisp has a strange syntax as that Lisp has no syntax. You write programs in the parse trees that get generated within the compiler when other languages are parsed. But these parse trees are fully accessible to your programs. You can write programs that manipulate them. In Lisp, these programs are called macros. They are programs that write programs.  
  
Programs that write programs? When would you ever want to do that? Not very often, if you think in Cobol. All the time, if you think in Lisp. It would be convenient here if I could give an example of a powerful macro, and say there! how about that? But if I did, it would just look like gibberish to someone who didn't know Lisp; there isn't room here to explain everything you'd need to know to understand what it meant. In [Ansi Common Lisp](acl.html) I tried to move things along as fast as I could, and even so I didn't get to macros until page 160.  
  
But I think I can give a kind of argument that might be convincing. The source code of the Viaweb editor was probably about 20-25% macros. Macros are harder to write than ordinary Lisp functions, and it's considered to be bad style to use them when they're not necessary. So every macro in that code is there because it has to be. What that means is that at least 20-25% of the code in this program is doing things that you can't easily do in any other language. However skeptical the Blub programmer might be about my claims for the mysterious powers of Lisp, this ought to make him curious. We weren't writing this code for our own amusement. We were a tiny startup, programming as hard as we could in order to put technical barriers between us and our competitors.  
  
A suspicious person might begin to wonder if there was some correlation here. A big chunk of our code was doing things that are very hard to do in other languages. The resulting software did things our competitors' software couldn't do. Maybe there was some kind of connection. I encourage you to follow that thread. There may be more to that old man hobbling along on his crutches than meets the eye.  
  
**Aikido for Startups**  
  
But I don't expect to convince anyone ([over 25](http://www.trollope.org/scheme.html)) to go out and learn Lisp. The purpose of this article is not to change anyone's mind, but to reassure people already interested in using Lisp-- people who know that Lisp is a powerful language, but worry because it isn't widely used. In a competitive situation, that's an advantage. Lisp's power is multiplied by the fact that your competitors don't get it.  
  
If you think of using Lisp in a startup, you shouldn't worry that it isn't widely understood. You should hope that it stays that way. And it's likely to. It's the nature of programming languages to make most people satisfied with whatever they currently use. Computer hardware changes so much faster than personal habits that programming practice is usually ten to twenty years behind the processor. At places like MIT they were writing programs in high-level languages in the early 1960s, but many companies continued to write code in machine language well into the 1980s. I bet a lot of people continued to write machine language until the processor, like a bartender eager to close up and go home, finally kicked them out by switching to a risc instruction set.  
  
Ordinarily technology changes fast. But programming languages are different: programming languages are not just technology, but what programmers think in. They're half technology and half religion.\[6\] And so the median language, meaning whatever language the median programmer uses, moves as slow as an iceberg. Garbage collection, introduced by Lisp in about 1960, is now widely considered to be a good thing. Runtime typing, ditto, is growing in popularity. Lexical closures, introduced by Lisp in the early 1970s, are now, just barely, on the radar screen. Macros, introduced by Lisp in the mid 1960s, are still terra incognita.  
  
Obviously, the median language has enormous momentum. I'm not proposing that you can fight this powerful force. What I'm proposing is exactly the opposite: that, like a practitioner of Aikido, you can use it against your opponents.  
  
If you work for a big company, this may not be easy. You will have a hard time convincing the pointy-haired boss to let you build things in Lisp, when he has just read in the paper that some other language is poised, like Ada was twenty years ago, to take over the world. But if you work for a startup that doesn't have pointy-haired bosses yet, you can, like we did, turn the Blub paradox to your advantage: you can use technology that your competitors, glued immovably to the median language, will never be able to match.  
  
If you ever do find yourself working for a startup, here's a handy tip for evaluating competitors. Read their job listings. Everything else on their site may be stock photos or the prose equivalent, but the job listings have to be specific about what they want, or they'll get the wrong candidates.  
  
During the years we worked on Viaweb I read a lot of job descriptions. A new competitor seemed to emerge out of the woodwork every month or so. The first thing I would do, after checking to see if they had a live online demo, was look at their job listings. After a couple years of this I could tell which companies to worry about and which not to. The more of an IT flavor the job descriptions had, the less dangerous the company was. The safest kind were the ones that wanted Oracle experience. You never had to worry about those. You were also safe if they said they wanted C++ or Java developers. If they wanted Perl or Python programmers, that would be a bit frightening-- that's starting to sound like a company where the technical side, at least, is run by real hackers. If I had ever seen a job posting looking for Lisp hackers, I would have been really worried.  
  
  
  
**Notes**  
  
\[1\] Viaweb at first had two parts: the editor, written in Lisp, which people used to build their sites, and the ordering system, written in C, which handled orders. The first version was mostly Lisp, because the ordering system was small. Later we added two more modules, an image generator written in C, and a back-office manager written mostly in Perl.  
  
In January 2003, Yahoo released a new version of the editor written in C++ and Perl. It's hard to say whether the program is no longer written in Lisp, though, because to translate this program into C++ they literally had to write a Lisp interpreter: the source files of all the page-generating templates are still, as far as I know, Lisp code. (See [Greenspun's Tenth Rule](quotes.html).)  
  
\[2\] Robert Morris says that I didn't need to be secretive, because even if our competitors had known we were using Lisp, they wouldn't have understood why: "If they were that smart they'd already be programming in Lisp."  
  
\[3\] All languages are equally powerful in the sense of being Turing equivalent, but that's not the sense of the word programmers care about. (No one wants to program a Turing machine.) The kind of power programmers care about may not be formally definable, but one way to explain it would be to say that it refers to features you could only get in the less powerful language by writing an interpreter for the more powerful language in it. If language A has an operator for removing spaces from strings and language B doesn't, that probably doesn't make A more powerful, because you can probably write a subroutine to do it in B. But if A supports, say, recursion, and B doesn't, that's not likely to be something you can fix by writing library functions.  
  
\[4\] Note to nerds: or possibly a lattice, narrowing toward the top; it's not the shape that matters here but the idea that there is at least a partial order.  
  
\[5\] It is a bit misleading to treat macros as a separate feature. In practice their usefulness is greatly enhanced by other Lisp features like lexical closures and rest parameters.  
  
\[6\] As a result, comparisons of programming languages either take the form of religious wars or undergraduate textbooks so determinedly neutral that they're really works of anthropology. People who value their peace, or want tenure, avoid the topic. But the question is only half a religious one; there is something there worth studying, especially if you want to design new languages.  
  

  

[More Technical Details](lwba.html)  
  

[Japanese Translation](http://www.shiro.dreamhost.com/scheme/trans/beating-the-averages-j.html)  
  

[Turkish Translation](http://ileriseviye.org/arasayfa.php?inode=beatingaverages-tr.html)  
  

[Uzbek Translation](http://bobir.livejournal.com/17358.html)  
  

[Orbitz Uses Lisp Too](carl.html)  
  

[How To Become A Hacker](http://www.catb.org/~esr/faqs/hacker-howto.html)  
  

[A Scheme Story](http://www.trollope.org/scheme.html)  
  

[Italian Translation](https://marcotrombetti.com/media)  
  

  

  
  

* * *

You'll find this essay and 14 others in [**_Hackers & Painters_**](http://www.amazon.com/gp/product/0596006624).



# Java's Cover

[](index.html)  
  

  
  
April 2001  
  
This essay developed out of conversations I've had with several other programmers about why Java smelled suspicious. It's not a critique of Java! It is a case study of hacker's radar.  
  
Over time, hackers develop a nose for good (and bad) technology. I thought it might be interesting to try and write down what made Java seem suspect to me.  
  
Some people who've read this think it's an interesting attempt to write about something that hasn't been written about before. Others say I will get in trouble for appearing to be writing about things I don't understand. So, just in case it does any good, let me clarify that I'm not writing here about Java (which I have never used) but about hacker's radar (which I have thought about a lot).  
  

* * *

  
  
The aphorism "you can't tell a book by its cover" originated in the times when books were sold in plain cardboard covers, to be bound by each purchaser according to his own taste. In those days, you couldn't tell a book by its cover. But publishing has advanced since then: present-day publishers work hard to make the cover something you can tell a book by.  
  
I spend a lot of time in bookshops and I feel as if I have by now learned to understand everything publishers mean to tell me about a book, and perhaps a bit more. The time I haven't spent in bookshops I've spent mostly in front of computers, and I feel as if I've learned, to some degree, to judge technology by its cover as well. It may be just luck, but I've saved myself from a few technologies that turned out to be real stinkers.  
  
So far, Java seems like a stinker to me. I've never written a Java program, never more than glanced over reference books about it, but I have a hunch that it won't be a very successful language. I may turn out to be mistaken; making predictions about technology is a dangerous business. But for what it's worth, as a sort of time capsule, here's why I don't like the look of Java:  
  
1\. It has been so energetically hyped. Real standards don't have to be promoted. No one had to promote C, or Unix, or HTML. A real standard tends to be already established by the time most people hear about it. On the hacker radar screen, Perl is as big as Java, or bigger, just on the strength of its own merits.  
  
2\. It's aimed low. In the original Java white paper, Gosling explicitly says Java was designed not to be too difficult for programmers used to C. It was designed to be another C++: C plus a few ideas taken from more advanced languages. Like the creators of sitcoms or junk food or package tours, Java's designers were consciously designing a product for people not as smart as them. Historically, languages designed for other people to use have been bad: Cobol, PL/I, Pascal, Ada, C++. The good languages have been those that were designed for their own creators: C, Perl, Smalltalk, Lisp.  
  
3\. It has ulterior motives. Someone once said that the world would be a better place if people only wrote books because they had something to say, rather than because they wanted to write a book. Likewise, the reason we hear about Java all the time is not because it has something to say about programming languages. We hear about Java as part of a plan by Sun to undermine Microsoft.  
  
4\. No one loves it. C, Perl, Python, Smalltalk, and Lisp programmers love their languages. I've never heard anyone say that they loved Java.  
  
5\. People are forced to use it. A lot of the people I know using Java are using it because they feel they have to. Either it's something they felt they had to do to get funded, or something they thought customers would want, or something they were told to do by management. These are smart people; if the technology was good, they'd have used it voluntarily.  
  
6\. It has too many cooks. The best programming languages have been developed by small groups. Java seems to be run by a committee. If it turns out to be a good language, it will be the first time in history that a committee has designed a good language.  
  
7\. It's bureaucratic. From what little I know about Java, there seem to be a lot of protocols for doing things. Really good languages aren't like that. They let you do what you want and get out of the way.  
  
8\. It's pseudo-hip. Sun now pretends that Java is a grassroots, open-source language effort like Perl or Python. This one just happens to be controlled by a giant company. So the language is likely to have the same drab clunkiness as anything else that comes out of a big company.  
  
9\. It's designed for large organizations. Large organizations have different aims from hackers. They want languages that are (believed to be) suitable for use by large teams of mediocre programmers-- languages with features that, like the speed limiters in U-Haul trucks, prevent fools from doing too much damage. Hackers don't like a language that talks down to them. Hackers just want power. Historically, languages designed for large organizations (PL/I, Ada) have lost, while hacker languages (C, Perl) have won. The reason: today's teenage hacker is tomorrow's CTO.  
  
10\. The wrong people like it. The programmers I admire most are not, on the whole, captivated by Java. Who does like Java? Suits, who don't know one language from another, but know that they keep hearing about Java in the press; programmers at big companies, who are amazed to find that there is something even better than C++; and plug-and-chug undergrads, who are ready to like anything that might get them a job (will this be on the test?). These people's opinions change with every wind.  
  
11\. Its daddy is in a pinch. Sun's business model is being undermined on two fronts. Cheap Intel processors, of the same type used in desktop machines, are now more than fast enough for servers. And FreeBSD seems to be at least as good an OS for servers as Solaris. Sun's advertising implies that you need Sun servers for industrial strength applications. If this were true, Yahoo would be first in line to buy Suns; but when I worked there, the servers were all Intel boxes running FreeBSD. This bodes ill for Sun's future. If Sun runs into trouble, they could drag Java down with them.  
  
12\. The DoD likes it. The Defense Department is encouraging developers to use Java. This seems to me the most damning sign of all. The Defense Department does a fine (though expensive) job of defending the country, but they love plans and procedures and protocols. Their culture is the opposite of hacker culture; on questions of software they will tend to bet wrong. The last time the DoD really liked a programming language, it was Ada.  
  
Bear in mind, this is not a critique of Java, but a critique of its cover. I don't know Java well enough to like it or dislike it. This is just an explanation of why I don't find that I'm eager to learn it.  
  
It may seem cavalier to dismiss a language before you've even tried writing programs in it. But this is something all programmers have to do. There are too many technologies out there to learn them all. You have to learn to judge by outward signs which will be worth your time. I have likewise cavalierly dismissed Cobol, Ada, Visual Basic, the IBM AS400, VRML, ISO 9000, the SET protocol, VMS, Novell Netware, and CORBA, among others. They just smelled wrong.  
  
It could be that in Java's case I'm mistaken. It could be that a language promoted by one big company to undermine another, designed by a committee for a "mainstream" audience, hyped to the skies, and beloved of the DoD, happens nonetheless to be a clean, beautiful, powerful language that I would love programming in. It could be, but it seems very unlikely.  
  

  

[Trevor Re: Java's Cover](trevrejavcov.html)  
  

[Berners-Lee Re: Java](bljava.html)  
  

[Being Popular](popular.html)  
  

[Sun Internal Memo](http://www.archub.org/javamemo.txt)  
  

[2005: BusinessWeek Agrees](http://www.businessweek.com/technology/content/dec2005/tc20051213_042973.htm)  
  

[Japanese Translation](http://d.hatena.ne.jp/lionfan/20070211)  
  

  

  
  

* * *



# Being Popular

[](index.html)  
  

  
  
May 2001  
  
_(This article was written as a kind of business plan for a [new language](arc.html). So it is missing (because it takes for granted) the most important feature of a good programming language: very powerful abstractions.)_  
  
A friend of mine once told an eminent operating systems expert that he wanted to design a really good programming language. The expert told him that it would be a waste of time, that programming languages don't become popular or unpopular based on their merits, and so no matter how good his language was, no one would use it. At least, that was what had happened to the language _he_ had designed.  
  
What does make a language popular? Do popular languages deserve their popularity? Is it worth trying to define a good programming language? How would you do it?  
  
I think the answers to these questions can be found by looking at hackers, and learning what they want. Programming languages are _for_ hackers, and a programming language is good as a programming language (rather than, say, an exercise in denotational semantics or compiler design) if and only if hackers like it.  
  
**1 The Mechanics of Popularity**  
  
It's true, certainly, that most people don't choose programming languages simply based on their merits. Most programmers are told what language to use by someone else. And yet I think the effect of such external factors on the popularity of programming languages is not as great as it's sometimes thought to be. I think a bigger problem is that a hacker's idea of a good programming language is not the same as most language designers'.  
  
Between the two, the hacker's opinion is the one that matters. Programming languages are not theorems. They're tools, designed for people, and they have to be designed to suit human strengths and weaknesses as much as shoes have to be designed for human feet. If a shoe pinches when you put it on, it's a bad shoe, however elegant it may be as a piece of sculpture.  
  
It may be that the majority of programmers can't tell a good language from a bad one. But that's no different with any other tool. It doesn't mean that it's a waste of time to try designing a good language. [Expert hackers](design.html) can tell a good language when they see one, and they'll use it. Expert hackers are a tiny minority, admittedly, but that tiny minority write all the good software, and their influence is such that the rest of the programmers will tend to use whatever language they use. Often, indeed, it is not merely influence but command: often the expert hackers are the very people who, as their bosses or faculty advisors, tell the other programmers what language to use.  
  
The opinion of expert hackers is not the only force that determines the relative popularity of programming languages — legacy software (Cobol) and hype (Ada, Java) also play a role — but I think it is the most powerful force over the long term. Given an initial critical mass and enough time, a programming language probably becomes about as popular as it deserves to be. And popularity further separates good languages from bad ones, because feedback from real live users always leads to improvements. Look at how much any popular language has changed during its life. Perl and Fortran are extreme cases, but even Lisp has changed a lot. Lisp 1.5 didn't have macros, for example; these evolved later, after hackers at MIT had spent a couple years using Lisp to write real programs. \[1\]  
  
So whether or not a language has to be good to be popular, I think a language has to be popular to be good. And it has to stay popular to stay good. The state of the art in programming languages doesn't stand still. And yet the Lisps we have today are still pretty much what they had at MIT in the mid-1980s, because that's the last time Lisp had a sufficiently large and demanding user base.  
  
Of course, hackers have to know about a language before they can use it. How are they to hear? From other hackers. But there has to be some initial group of hackers using the language for others even to hear about it. I wonder how large this group has to be; how many users make a critical mass? Off the top of my head, I'd say twenty. If a language had twenty separate users, meaning twenty users who decided on their own to use it, I'd consider it to be real.  
  
Getting there can't be easy. I would not be surprised if it is harder to get from zero to twenty than from twenty to a thousand. The best way to get those initial twenty users is probably to use a trojan horse: to give people an application they want, which happens to be written in the new language.  
  
**2 External Factors**  
  
Let's start by acknowledging one external factor that does affect the popularity of a programming language. To become popular, a programming language has to be the scripting language of a popular system. Fortran and Cobol were the scripting languages of early IBM mainframes. C was the scripting language of Unix, and so, later, was Perl. Tcl is the scripting language of Tk. Java and Javascript are intended to be the scripting languages of web browsers.  
  
Lisp is not a massively popular language because it is not the scripting language of a massively popular system. What popularity it retains dates back to the 1960s and 1970s, when it was the scripting language of MIT. A lot of the great programmers of the day were associated with MIT at some point. And in the early 1970s, before C, MIT's dialect of Lisp, called MacLisp, was one of the only programming languages a serious hacker would want to use.  
  
Today Lisp is the scripting language of two moderately popular systems, Emacs and Autocad, and for that reason I suspect that most of the Lisp programming done today is done in Emacs Lisp or AutoLisp.  
  
Programming languages don't exist in isolation. To hack is a transitive verb — hackers are usually hacking something — and in practice languages are judged relative to whatever they're used to hack. So if you want to design a popular language, you either have to supply more than a language, or you have to design your language to replace the scripting language of some existing system.  
  
Common Lisp is unpopular partly because it's an orphan. It did originally come with a system to hack: the Lisp Machine. But Lisp Machines (along with parallel computers) were steamrollered by the increasing power of general purpose processors in the 1980s. Common Lisp might have remained popular if it had been a good scripting language for Unix. It is, alas, an atrociously bad one.  
  
One way to describe this situation is to say that a language isn't judged on its own merits. Another view is that a programming language really isn't a programming language unless it's also the scripting language of something. This only seems unfair if it comes as a surprise. I think it's no more unfair than expecting a programming language to have, say, an implementation. It's just part of what a programming language is.  
  
A programming language does need a good implementation, of course, and this must be free. Companies will pay for software, but individual hackers won't, and it's the hackers you need to attract.  
  
A language also needs to have a book about it. The book should be thin, well-written, and full of good examples. K&R is the ideal here. At the moment I'd almost say that a language has to have a book published by O'Reilly. That's becoming the test of mattering to hackers.  
  
There should be online documentation as well. In fact, the book can start as online documentation. But I don't think that physical books are outmoded yet. Their format is convenient, and the de facto censorship imposed by publishers is a useful if imperfect filter. Bookstores are one of the most important places for learning about new languages.  
  
**3 Brevity**  
  
Given that you can supply the three things any language needs — a free implementation, a book, and something to hack — how do you make a language that hackers will like?  
  
One thing hackers like is brevity. Hackers are lazy, in the same way that mathematicians and modernist architects are lazy: they hate anything extraneous. It would not be far from the truth to say that a hacker about to write a program decides what language to use, at least subconsciously, based on the total number of characters he'll have to type. If this isn't precisely how hackers think, a language designer would do well to act as if it were.  
  
It is a mistake to try to baby the user with long-winded expressions that are meant to resemble English. Cobol is notorious for this flaw. A hacker would consider being asked to write  
  
add x to y giving z  
  
instead of  
  
z = x+y  
  
as something between an insult to his intelligence and a sin against God.  
  
It has sometimes been said that Lisp should use first and rest instead of car and cdr, because it would make programs easier to read. Maybe for the first couple hours. But a hacker can learn quickly enough that car means the first element of a list and cdr means the rest. Using first and rest means 50% more typing. And they are also different lengths, meaning that the arguments won't line up when they're called, as car and cdr often are, in successive lines. I've found that it matters a lot how code lines up on the page. I can barely read Lisp code when it is set in a variable-width font, and friends say this is true for other languages too.  
  
Brevity is one place where strongly typed languages lose. All other things being equal, no one wants to begin a program with a bunch of declarations. Anything that can be implicit, should be.  
  
The individual tokens should be short as well. Perl and Common Lisp occupy opposite poles on this question. Perl programs can be almost cryptically dense, while the names of built-in Common Lisp operators are comically long. The designers of Common Lisp probably expected users to have text editors that would type these long names for them. But the cost of a long name is not just the cost of typing it. There is also the cost of reading it, and the cost of the space it takes up on your screen.  
  
**4 Hackability**  
  
There is one thing more important than brevity to a hacker: being able to do what you want. In the history of programming languages a surprising amount of effort has gone into preventing programmers from doing things considered to be improper. This is a dangerously presumptuous plan. How can the language designer know what the programmer is going to need to do? I think language designers would do better to consider their target user to be a genius who will need to do things they never anticipated, rather than a bumbler who needs to be protected from himself. The bumbler will shoot himself in the foot anyway. You may save him from referring to variables in another package, but you can't save him from writing a badly designed program to solve the wrong problem, and taking forever to do it.  
  
Good programmers often want to do dangerous and unsavory things. By unsavory I mean things that go behind whatever semantic facade the language is trying to present: getting hold of the internal representation of some high-level abstraction, for example. Hackers like to hack, and hacking means getting inside things and second guessing the original designer.  
  
_Let yourself be second guessed._ When you make any tool, people use it in ways you didn't intend, and this is especially true of a highly articulated tool like a programming language. Many a hacker will want to tweak your semantic model in a way that you never imagined. I say, let them; give the programmer access to as much internal stuff as you can without endangering runtime systems like the garbage collector.  
  
In Common Lisp I have often wanted to iterate through the fields of a struct — to comb out references to a deleted object, for example, or find fields that are uninitialized. I know the structs are just vectors underneath. And yet I can't write a general purpose function that I can call on any struct. I can only access the fields by name, because that's what a struct is supposed to mean.  
  
A hacker may only want to subvert the intended model of things once or twice in a big program. But what a difference it makes to be able to. And it may be more than a question of just solving a problem. There is a kind of pleasure here too. Hackers share the surgeon's secret pleasure in poking about in gross innards, the teenager's secret pleasure in popping zits. \[2\] For boys, at least, certain kinds of horrors are fascinating. Maxim magazine publishes an annual volume of photographs, containing a mix of pin-ups and grisly accidents. They know their audience.  
  
Historically, Lisp has been good at letting hackers have their way. The political correctness of Common Lisp is an aberration. Early Lisps let you get your hands on everything. A good deal of that spirit is, fortunately, preserved in macros. What a wonderful thing, to be able to make arbitrary transformations on the source code.  
  
Classic macros are a real hacker's tool — simple, powerful, and dangerous. It's so easy to understand what they do: you call a function on the macro's arguments, and whatever it returns gets inserted in place of the macro call. Hygienic macros embody the opposite principle. They try to protect you from understanding what they're doing. I have never heard hygienic macros explained in one sentence. And they are a classic example of the dangers of deciding what programmers are allowed to want. Hygienic macros are intended to protect me from variable capture, among other things, but variable capture is exactly what I want in some macros.  
  
A really good language should be both clean and dirty: cleanly designed, with a small core of well understood and highly orthogonal operators, but dirty in the sense that it lets hackers have their way with it. C is like this. So were the early Lisps. A real hacker's language will always have a slightly raffish character.  
  
A good programming language should have features that make the kind of people who use the phrase "software engineering" shake their heads disapprovingly. At the other end of the continuum are languages like Ada and Pascal, models of propriety that are good for teaching and not much else.  
  
**5 Throwaway Programs**  
  
To be attractive to hackers, a language must be good for writing the kinds of programs they want to write. And that means, perhaps surprisingly, that it has to be good for writing throwaway programs.  
  
A throwaway program is a program you write quickly for some limited task: a program to automate some system administration task, or generate test data for a simulation, or convert data from one format to another. The surprising thing about throwaway programs is that, like the "temporary" buildings built at so many American universities during World War II, they often don't get thrown away. Many evolve into real programs, with real features and real users.  
  
I have a hunch that the best big programs begin life this way, rather than being designed big from the start, like the Hoover Dam. It's terrifying to build something big from scratch. When people take on a project that's too big, they become overwhelmed. The project either gets bogged down, or the result is sterile and wooden: a shopping mall rather than a real downtown, Brasilia rather than Rome, Ada rather than C.  
  
Another way to get a big program is to start with a throwaway program and keep improving it. This approach is less daunting, and the design of the program benefits from evolution. I think, if one looked, that this would turn out to be the way most big programs were developed. And those that did evolve this way are probably still written in whatever language they were first written in, because it's rare for a program to be ported, except for political reasons. And so, paradoxically, if you want to make a language that is used for big systems, you have to make it good for writing throwaway programs, because that's where big systems come from.  
  
Perl is a striking example of this idea. It was not only designed for writing throwaway programs, but was pretty much a throwaway program itself. Perl began life as a collection of utilities for generating reports, and only evolved into a programming language as the throwaway programs people wrote in it grew larger. It was not until Perl 5 (if then) that the language was suitable for writing serious programs, and yet it was already massively popular.  
  
What makes a language good for throwaway programs? To start with, it must be readily available. A throwaway program is something that you expect to write in an hour. So the language probably must already be installed on the computer you're using. It can't be something you have to install before you use it. It has to be there. C was there because it came with the operating system. Perl was there because it was originally a tool for system administrators, and yours had already installed it.  
  
Being available means more than being installed, though. An interactive language, with a command-line interface, is more available than one that you have to compile and run separately. A popular programming language should be interactive, and start up fast.  
  
Another thing you want in a throwaway program is brevity. Brevity is always attractive to hackers, and never more so than in a program they expect to turn out in an hour.  
  
**6 Libraries**  
  
Of course the ultimate in brevity is to have the program already written for you, and merely to call it. And this brings us to what I think will be an increasingly important feature of programming languages: library functions. Perl wins because it has large libraries for manipulating strings. This class of library functions are especially important for throwaway programs, which are often originally written for converting or extracting data. Many Perl programs probably begin as just a couple library calls stuck together.  
  
I think a lot of the advances that happen in programming languages in the next fifty years will have to do with library functions. I think future programming languages will have libraries that are as carefully designed as the core language. Programming language design will not be about whether to make your language strongly or weakly typed, or object oriented, or functional, or whatever, but about how to design great libraries. The kind of language designers who like to think about how to design type systems may shudder at this. It's almost like writing applications! Too bad. Languages are for programmers, and libraries are what programmers need.  
  
It's hard to design good libraries. It's not simply a matter of writing a lot of code. Once the libraries get too big, it can sometimes take longer to find the function you need than to write the code yourself. Libraries need to be designed using a small set of orthogonal operators, just like the core language. It ought to be possible for the programmer to guess what library call will do what he needs.  
  
Libraries are one place Common Lisp falls short. There are only rudimentary libraries for manipulating strings, and almost none for talking to the operating system. For historical reasons, Common Lisp tries to pretend that the OS doesn't exist. And because you can't talk to the OS, you're unlikely to be able to write a serious program using only the built-in operators in Common Lisp. You have to use some implementation-specific hacks as well, and in practice these tend not to give you everything you want. Hackers would think a lot more highly of Lisp if Common Lisp had powerful string libraries and good OS support.  
  
**7 Syntax**  
  
Could a language with Lisp's syntax, or more precisely, lack of syntax, ever become popular? I don't know the answer to this question. I do think that syntax is not the main reason Lisp isn't currently popular. Common Lisp has worse problems than unfamiliar syntax. I know several programmers who are comfortable with prefix syntax and yet use Perl by default, because it has powerful string libraries and can talk to the os.  
  
There are two possible problems with prefix notation: that it is unfamiliar to programmers, and that it is not dense enough. The conventional wisdom in the Lisp world is that the first problem is the real one. I'm not so sure. Yes, prefix notation makes ordinary programmers panic. But I don't think ordinary programmers' opinions matter. Languages become popular or unpopular based on what expert hackers think of them, and I think expert hackers might be able to deal with prefix notation. Perl syntax can be pretty incomprehensible, but that has not stood in the way of Perl's popularity. If anything it may have helped foster a Perl cult.  
  
A more serious problem is the diffuseness of prefix notation. For expert hackers, that really is a problem. No one wants to write (aref a x y) when they could write a\[x,y\].  
  
In this particular case there is a way to finesse our way out of the problem. If we treat data structures as if they were functions on indexes, we could write (a x y) instead, which is even shorter than the Perl form. Similar tricks may shorten other types of expressions.  
  
We can get rid of (or make optional) a lot of parentheses by making indentation significant. That's how programmers read code anyway: when indentation says one thing and delimiters say another, we go by the indentation. Treating indentation as significant would eliminate this common source of bugs as well as making programs shorter.  
  
Sometimes infix syntax is easier to read. This is especially true for math expressions. I've used Lisp my whole programming life and I still don't find prefix math expressions natural. And yet it is convenient, especially when you're generating code, to have operators that take any number of arguments. So if we do have infix syntax, it should probably be implemented as some kind of read-macro.  
  
I don't think we should be religiously opposed to introducing syntax into Lisp, as long as it translates in a well-understood way into underlying s-expressions. There is already a good deal of syntax in Lisp. It's not necessarily bad to introduce more, as long as no one is forced to use it. In Common Lisp, some delimiters are reserved for the language, suggesting that at least some of the designers intended to have more syntax in the future.  
  
One of the most egregiously unlispy pieces of syntax in Common Lisp occurs in format strings; format is a language in its own right, and that language is not Lisp. If there were a plan for introducing more syntax into Lisp, format specifiers might be able to be included in it. It would be a good thing if macros could generate format specifiers the way they generate any other kind of code.  
  
An eminent Lisp hacker told me that his copy of CLTL falls open to the section format. Mine too. This probably indicates room for improvement. It may also mean that programs do a lot of I/O.  
  
**8 Efficiency**  
  
A good language, as everyone knows, should generate fast code. But in practice I don't think fast code comes primarily from things you do in the design of the language. As Knuth pointed out long ago, speed only matters in certain critical bottlenecks. And as many programmers have observed since, one is very often mistaken about where these bottlenecks are.  
  
So, in practice, the way to get fast code is to have a very good profiler, rather than by, say, making the language strongly typed. You don't need to know the type of every argument in every call in the program. You do need to be able to declare the types of arguments in the bottlenecks. And even more, you need to be able to find out where the bottlenecks are.  
  
One complaint people have had with Lisp is that it's hard to tell what's expensive. This might be true. It might also be inevitable, if you want to have a very abstract language. And in any case I think good profiling would go a long way toward fixing the problem: you'd soon learn what was expensive.  
  
Part of the problem here is social. Language designers like to write fast compilers. That's how they measure their skill. They think of the profiler as an add-on, at best. But in practice a good profiler may do more to improve the speed of actual programs written in the language than a compiler that generates fast code. Here, again, language designers are somewhat out of touch with their users. They do a really good job of solving slightly the wrong problem.  
  
It might be a good idea to have an active profiler — to push performance data to the programmer instead of waiting for him to come asking for it. For example, the editor could display bottlenecks in red when the programmer edits the source code. Another approach would be to somehow represent what's happening in running programs. This would be an especially big win in server-based applications, where you have lots of running programs to look at. An active profiler could show graphically what's happening in memory as a program's running, or even make sounds that tell what's happening.  
  
Sound is a good cue to problems. In one place I worked, we had a big board of dials showing what was happening to our web servers. The hands were moved by little servomotors that made a slight noise when they turned. I couldn't see the board from my desk, but I found that I could tell immediately, by the sound, when there was a problem with a server.  
  
It might even be possible to write a profiler that would automatically detect inefficient algorithms. I would not be surprised if certain patterns of memory access turned out to be sure signs of bad algorithms. If there were a little guy running around inside the computer executing our programs, he would probably have as long and plaintive a tale to tell about his job as a federal government employee. I often have a feeling that I'm sending the processor on a lot of wild goose chases, but I've never had a good way to look at what it's doing.  
  
A number of Lisps now compile into byte code, which is then executed by an interpreter. This is usually done to make the implementation easier to port, but it could be a useful language feature. It might be a good idea to make the byte code an official part of the language, and to allow programmers to use inline byte code in bottlenecks. Then such optimizations would be portable too.  
  
The nature of speed, as perceived by the end-user, may be changing. With the rise of server-based applications, more and more programs may turn out to be i/o-bound. It will be worth making i/o fast. The language can help with straightforward measures like simple, fast, formatted output functions, and also with deep structural changes like caching and persistent objects.  
  
Users are interested in response time. But another kind of efficiency will be increasingly important: the number of simultaneous users you can support per processor. Many of the interesting applications written in the near future will be server-based, and the number of users per server is the critical question for anyone hosting such applications. In the capital cost of a business offering a server-based application, this is the divisor.  
  
For years, efficiency hasn't mattered much in most end-user applications. Developers have been able to assume that each user would have an increasingly powerful processor sitting on their desk. And by Parkinson's Law, software has expanded to use the resources available. That will change with server-based applications. In that world, the hardware and software will be supplied together. For companies that offer server-based applications, it will make a very big difference to the bottom line how many users they can support per server.  
  
In some applications, the processor will be the limiting factor, and execution speed will be the most important thing to optimize. But often memory will be the limit; the number of simultaneous users will be determined by the amount of memory you need for each user's data. The language can help here too. Good support for threads will enable all the users to share a single heap. It may also help to have persistent objects and/or language level support for lazy loading.  
  
**9 Time**  
  
The last ingredient a popular language needs is time. No one wants to write programs in a language that might go away, as so many programming languages do. So most hackers will tend to wait until a language has been around for a couple years before even considering using it.  
  
Inventors of wonderful new things are often surprised to discover this, but you need time to get any message through to people. A friend of mine rarely does anything the first time someone asks him. He knows that people sometimes ask for things that they turn out not to want. To avoid wasting his time, he waits till the third or fourth time he's asked to do something; by then, whoever's asking him may be fairly annoyed, but at least they probably really do want whatever they're asking for.  
  
Most people have learned to do a similar sort of filtering on new things they hear about. They don't even start paying attention until they've heard about something ten times. They're perfectly justified: the majority of hot new whatevers do turn out to be a waste of time, and eventually go away. By delaying learning VRML, I avoided having to learn it at all.  
  
So anyone who invents something new has to expect to keep repeating their message for years before people will start to get it. We wrote what was, as far as I know, the first web-server based application, and it took us years to get it through to people that it didn't have to be downloaded. It wasn't that they were stupid. They just had us tuned out.  
  
The good news is, simple repetition solves the problem. All you have to do is keep telling your story, and eventually people will start to hear. It's not when people notice you're there that they pay attention; it's when they notice you're still there.  
  
It's just as well that it usually takes a while to gain momentum. Most technologies evolve a good deal even after they're first launched — programming languages especially. Nothing could be better, for a new techology, than a few years of being used only by a small number of early adopters. Early adopters are sophisticated and demanding, and quickly flush out whatever flaws remain in your technology. When you only have a few users you can be in close contact with all of them. And early adopters are forgiving when you improve your system, even if this causes some breakage.  
  
There are two ways new technology gets introduced: the organic growth method, and the big bang method. The organic growth method is exemplified by the classic seat-of-the-pants underfunded garage startup. A couple guys, working in obscurity, develop some new technology. They launch it with no marketing and initially have only a few (fanatically devoted) users. They continue to improve the technology, and meanwhile their user base grows by word of mouth. Before they know it, they're big.  
  
The other approach, the big bang method, is exemplified by the VC-backed, heavily marketed startup. They rush to develop a product, launch it with great publicity, and immediately (they hope) have a large user base.  
  
Generally, the garage guys envy the big bang guys. The big bang guys are smooth and confident and respected by the VCs. They can afford the best of everything, and the PR campaign surrounding the launch has the side effect of making them celebrities. The organic growth guys, sitting in their garage, feel poor and unloved. And yet I think they are often mistaken to feel sorry for themselves. Organic growth seems to yield better technology and richer founders than the big bang method. If you look at the dominant technologies today, you'll find that most of them grew organically.  
  
This pattern doesn't only apply to companies. You see it in sponsored research too. Multics and Common Lisp were big-bang projects, and Unix and MacLisp were organic growth projects.  
  
**10 Redesign**  
  
"The best writing is rewriting," wrote E. B. White. Every good writer knows this, and it's true for software too. The most important part of design is redesign. Programming languages, especially, don't get redesigned enough.  
  
To write good software you must simultaneously keep two opposing ideas in your head. You need the young hacker's naive faith in his abilities, and at the same time the veteran's skepticism. You have to be able to think [how hard can it be?](http://www.trevorblackwell.com) with one half of your brain while thinking [it will never work](http://www.pdos.lcs.mit.edu/~rtm/) with the other.  
  
The trick is to realize that there's no real contradiction here. You want to be optimistic and skeptical about two different things. You have to be optimistic about the possibility of solving the problem, but skeptical about the value of whatever solution you've got so far.  
  
People who do good work often think that whatever they're working on is no good. Others see what they've done and are full of wonder, but the creator is full of worry. This pattern is no coincidence: it is the worry that made the work good.  
  
If you can keep hope and worry balanced, they will drive a project forward the same way your two legs drive a bicycle forward. In the first phase of the two-cycle innovation engine, you work furiously on some problem, inspired by your confidence that you'll be able to solve it. In the second phase, you look at what you've done in the cold light of morning, and see all its flaws very clearly. But as long as your critical spirit doesn't outweigh your hope, you'll be able to look at your admittedly incomplete system, and think, how hard can it be to get the rest of the way?, thereby continuing the cycle.  
  
It's tricky to keep the two forces balanced. In young hackers, optimism predominates. They produce something, are convinced it's great, and never improve it. In old hackers, skepticism predominates, and they won't even dare to take on ambitious projects.  
  
Anything you can do to keep the redesign cycle going is good. Prose can be rewritten over and over until you're happy with it. But software, as a rule, doesn't get redesigned enough. Prose has readers, but software has _users._ If a writer rewrites an essay, people who read the old version are unlikely to complain that their thoughts have been broken by some newly introduced incompatibility.  
  
Users are a double-edged sword. They can help you improve your language, but they can also deter you from improving it. So choose your users carefully, and be slow to grow their number. Having users is like optimization: the wise course is to delay it. Also, as a general rule, you can at any given time get away with changing more than you think. Introducing change is like pulling off a bandage: the pain is a memory almost as soon as you feel it.  
  
Everyone knows that it's not a good idea to have a language designed by a committee. Committees yield bad design. But I think the worst danger of committees is that they interfere with redesign. It is so much work to introduce changes that no one wants to bother. Whatever a committee decides tends to stay that way, even if most of the members don't like it.  
  
Even a committee of two gets in the way of redesign. This happens particularly in the interfaces between pieces of software written by two different people. To change the interface both have to agree to change it at once. And so interfaces tend not to change at all, which is a problem because they tend to be one of the most ad hoc parts of any system.  
  
One solution here might be to design systems so that interfaces are horizontal instead of vertical — so that modules are always vertically stacked strata of abstraction. Then the interface will tend to be owned by one of them. The lower of two levels will either be a language in which the upper is written, in which case the lower level will own the interface, or it will be a slave, in which case the interface can be dictated by the upper level.  
  
**11 Lisp**  
  
What all this implies is that there is hope for a new Lisp. There is hope for any language that gives hackers what they want, including Lisp. I think we may have made a mistake in thinking that hackers are turned off by Lisp's strangeness. This comforting illusion may have prevented us from seeing the real problem with Lisp, or at least Common Lisp, which is that it sucks for doing what hackers want to do. A hacker's language needs powerful libraries and something to hack. Common Lisp has neither. A hacker's language is terse and hackable. Common Lisp is not.  
  
The good news is, it's not Lisp that sucks, but Common Lisp. If we can develop a new Lisp that is a real hacker's language, I think hackers will use it. They will use whatever language does the job. All we have to do is make sure this new Lisp does some important job better than other languages.  
  
History offers some encouragement. Over time, successive new programming languages have taken more and more features from Lisp. There is no longer much left to copy before the language you've made is Lisp. The latest hot language, Python, is a watered-down Lisp with infix syntax and no macros. A new Lisp would be a natural step in this progression.  
  
I sometimes think that it would be a good marketing trick to call it an improved version of Python. That sounds hipper than Lisp. To many people, Lisp is a slow AI language with a lot of parentheses. Fritz Kunze's official biography carefully avoids mentioning the L-word. But my guess is that we shouldn't be afraid to call the new Lisp Lisp. Lisp still has a lot of latent respect among the very best hackers — the ones who took 6.001 and understood it, for example. And those are the users you need to win.  
  
In "How to Become a Hacker," Eric Raymond describes Lisp as something like Latin or Greek — a language you should learn as an intellectual exercise, even though you won't actually use it:

> Lisp is worth learning for the profound enlightenment experience you will have when you finally get it; that experience will make you a better programmer for the rest of your days, even if you never actually use Lisp itself a lot.

If I didn't know Lisp, reading this would set me asking questions. A language that would make me a better programmer, if it means anything at all, means a language that would be better for programming. And that is in fact the implication of what Eric is saying.  
  
As long as that idea is still floating around, I think hackers will be receptive enough to a new Lisp, even if it is called Lisp. But this Lisp must be a hacker's language, like the classic Lisps of the 1970s. It must be terse, simple, and hackable. And it must have powerful libraries for doing what hackers want to do now.  
  
In the matter of libraries I think there is room to beat languages like Perl and Python at their own game. A lot of the new applications that will need to be written in the coming years will be [server-based applications](road.html). There's no reason a new Lisp shouldn't have string libraries as good as Perl, and if this new Lisp also had powerful libraries for server-based applications, it could be very popular. Real hackers won't turn up their noses at a new tool that will let them solve hard problems with a few library calls. Remember, hackers are lazy.  
  
It could be an even bigger win to have core language support for server-based applications. For example, explicit support for programs with multiple users, or data ownership at the level of type tags.  
  
Server-based applications also give us the answer to the question of what this new Lisp will be used to hack. It would not hurt to make Lisp better as a scripting language for Unix. (It would be hard to make it worse.) But I think there are areas where existing languages would be easier to beat. I think it might be better to follow the model of Tcl, and supply the Lisp together with a complete system for supporting server-based applications. Lisp is a natural fit for server-based applications. Lexical closures provide a way to get the effect of subroutines when the ui is just a series of web pages. S-expressions map nicely onto html, and macros are good at generating it. There need to be better tools for writing server-based applications, and there needs to be a new Lisp, and the two would work very well together.  
  
**12 The Dream Language**  
  
By way of summary, let's try describing the hacker's dream language. The dream language is [beautiful](taste.html), clean, and terse. It has an interactive toplevel that starts up fast. You can write programs to solve common problems with very little code. Nearly all the code in any program you write is code that's specific to your application. Everything else has been done for you.  
  
The syntax of the language is brief to a fault. You never have to type an unnecessary character, or even to use the shift key much.  
  
Using big abstractions you can write the first version of a program very quickly. Later, when you want to optimize, there's a really good profiler that tells you where to focus your attention. You can make inner loops blindingly fast, even writing inline byte code if you need to.  
  
There are lots of good examples to learn from, and the language is intuitive enough that you can learn how to use it from examples in a couple minutes. You don't need to look in the manual much. The manual is thin, and has few warnings and qualifications.  
  
The language has a small core, and powerful, highly orthogonal libraries that are as carefully designed as the core language. The libraries all work well together; everything in the language fits together like the parts in a fine camera. Nothing is deprecated, or retained for compatibility. The source code of all the libraries is readily available. It's easy to talk to the operating system and to applications written in other languages.  
  
The language is built in layers. The higher-level abstractions are built in a very transparent way out of lower-level abstractions, which you can get hold of if you want.  
  
Nothing is hidden from you that doesn't absolutely have to be. The language offers abstractions only as a way of saving you work, rather than as a way of telling you what to do. In fact, the language encourages you to be an equal participant in its design. You can change everything about it, including even its syntax, and anything you write has, as much as possible, the same status as what comes predefined.  
  
  
  
**Notes**  
  
\[1\] Macros very close to the modern idea were proposed by Timothy Hart in 1964, two years after Lisp 1.5 was released. What was missing, initially, were ways to avoid variable capture and multiple evaluation; Hart's examples are subject to both.  
  
\[2\] In _When the Air Hits Your Brain,_ neurosurgeon Frank Vertosick recounts a conversation in which his chief resident, Gary, talks about the difference between surgeons and internists ("fleas"):

> Gary and I ordered a large pizza and found an open booth. The chief lit a cigarette. "Look at those goddamn fleas, jabbering about some disease they'll see once in their lifetimes. That's the trouble with fleas, they only like the bizarre stuff. They hate their bread and butter cases. That's the difference between us and the fucking fleas. See, we love big juicy lumbar disc herniations, but they hate hypertension...."

It's hard to think of a lumbar disc herniation as juicy (except literally). And yet I think I know what they mean. I've often had a juicy bug to track down. Someone who's not a programmer would find it hard to imagine that there could be pleasure in a bug. Surely it's better if everything just works. In one way, it is. And yet there is undeniably a grim satisfaction in hunting down certain sorts of bugs.  
  

  

[Postscript Version](https://sep.yimg.com/ty/cdn/paulgraham/bepop.ps?t=1595850613&)  
  

[Arc](http://www.paulgraham.com/arc.html)  
  

[Five Questions about Language Design](langdes.html)  
  

[How to Become a Hacker](http://www.catb.org/~esr/faqs/hacker-howto.html)  
  

[Japanese Translation](http://www.shiro.dreamhost.com/scheme/trans/being-popular-j.html)  
  

  

  
  

* * *



# Five Questions about Language Design

[](index.html)  
  

  
  
May 2001

_(These are some notes I made for a panel discussion on programming language design at MIT on May 10, 2001.)_  
  
  
  
  
  
**1\. Programming Languages Are for People.**  
  
Programming languages are how people talk to computers. The computer would be just as happy speaking any language that was unambiguous. The reason we have high level languages is because people can't deal with machine language. The point of programming languages is to prevent our poor frail human brains from being overwhelmed by a mass of detail.  
  
Architects know that some kinds of design problems are more personal than others. One of the cleanest, most abstract design problems is designing bridges. There your job is largely a matter of spanning a given distance with the least material. The other end of the spectrum is designing chairs. Chair designers have to spend their time thinking about human butts.  
  
Software varies in the same way. Designing algorithms for routing data through a network is a nice, abstract problem, like designing bridges. Whereas designing programming languages is like designing chairs: it's all about dealing with human weaknesses.  
  
Most of us hate to acknowledge this. Designing systems of great mathematical elegance sounds a lot more appealing to most of us than pandering to human weaknesses. And there is a role for mathematical elegance: some kinds of elegance make programs easier to understand. But elegance is not an end in itself.  
  
And when I say languages have to be designed to suit human weaknesses, I don't mean that languages have to be designed for bad programmers. In fact I think you ought to design for the [best programmers](design.html), but even the best programmers have limitations. I don't think anyone would like programming in a language where all the variables were the letter x with integer subscripts.  
  
**2\. Design for Yourself and Your Friends.**  
  
If you look at the history of programming languages, a lot of the best ones were languages designed for their own authors to use, and a lot of the worst ones were designed for other people to use.  
  
When languages are designed for other people, it's always a specific group of other people: people not as smart as the language designer. So you get a language that talks down to you. Cobol is the most extreme case, but a lot of languages are pervaded by this spirit.  
  
It has nothing to do with how abstract the language is. C is pretty low-level, but it was designed for its authors to use, and that's why hackers like it.  
  
The argument for designing languages for bad programmers is that there are more bad programmers than good programmers. That may be so. But those few good programmers write a disproportionately large percentage of the software.  
  
I'm interested in the question, how do you design a language that the very best hackers will like? I happen to think this is identical to the question, how do you design a good programming language?, but even if it isn't, it is at least an interesting question.  
  
**3\. Give the Programmer as Much Control as Possible.**  
  
Many languages (especially the ones designed for other people) have the attitude of a governess: they try to prevent you from doing things that they think aren't good for you. I like the opposite approach: give the programmer as much control as you can.  
  
When I first learned Lisp, what I liked most about it was that it considered me an equal partner. In the other languages I had learned up till then, there was the language and there was my program, written in the language, and the two were very separate. But in Lisp the functions and macros I wrote were just like those that made up the language itself. I could rewrite the language if I wanted. It had the same appeal as open-source software.  
  
**4\. Aim for Brevity.**  
  
Brevity is underestimated and even scorned. But if you look into the hearts of hackers, you'll see that they really love it. How many times have you heard hackers speak fondly of how in, say, APL, they could do amazing things with just a couple lines of code? I think anything that really smart people really love is worth paying attention to.  
  
I think almost anything you can do to make programs shorter is good. There should be lots of library functions; anything that can be implicit should be; the syntax should be terse to a fault; even the names of things should be short.  
  
And it's not only programs that should be short. The manual should be thin as well. A good part of manuals is taken up with clarifications and reservations and warnings and special cases. If you force yourself to shorten the manual, in the best case you do it by fixing the things in the language that required so much explanation.  
  
**5\. Admit What Hacking Is.**  
  
A lot of people wish that hacking was mathematics, or at least something like a natural science. I think hacking is more like architecture. Architecture is related to physics, in the sense that architects have to design buildings that don't fall down, but the actual goal of architects is to make great buildings, not to make discoveries about statics.  
  
What hackers like to do is make great programs. And I think, at least in our own minds, we have to remember that it's an admirable thing to write great programs, even when this work doesn't translate easily into the conventional intellectual currency of research papers. Intellectually, it is just as worthwhile to design a language programmers will love as it is to design a horrible one that embodies some idea you can publish a paper about.  
  
  
  
  
  
**1\. How to Organize Big Libraries?**  
  
Libraries are becoming an increasingly important component of programming languages. They're also getting bigger, and this can be dangerous. If it takes longer to find the library function that will do what you want than it would take to write it yourself, then all that code is doing nothing but make your manual thick. (The Symbolics manuals were a case in point.) So I think we will have to work on ways to organize libraries. The ideal would be to design them so that the programmer could guess what library call would do the right thing.  
  
**2\. Are People Really Scared of Prefix Syntax?**  
  
This is an open problem in the sense that I have wondered about it for years and still don't know the answer. Prefix syntax seems perfectly natural to me, except possibly for math. But it could be that a lot of Lisp's unpopularity is simply due to having an unfamiliar syntax. Whether to do anything about it, if it is true, is another question.

**3\. What Do You Need for Server-Based Software?**

I think a lot of the most exciting new applications that get written in the next twenty years will be Web-based applications, meaning programs that sit on the server and talk to you through a Web browser. And to write these kinds of programs we may need some new things.  
  
One thing we'll need is support for the new way that server-based apps get released. Instead of having one or two big releases a year, like desktop software, server-based apps get released as a series of small changes. You may have as many as five or ten releases a day. And as a rule everyone will always use the latest version.  
  
You know how you can design programs to be debuggable? Well, server-based software likewise has to be designed to be changeable. You have to be able to change it easily, or at least to know what is a small change and what is a momentous one.  
  
Another thing that might turn out to be useful for server based software, surprisingly, is continuations. In Web-based software you can use something like continuation-passing style to get the effect of [subroutines](lwba.html) in the inherently stateless world of a Web session. Maybe it would be worthwhile having actual continuations, if it was not too expensive.  
  
**4\. What New Abstractions Are Left to Discover?**  
  
I'm not sure how reasonable a hope this is, but one thing I would really love to do, personally, is discover a new abstraction-- something that would make as much of a difference as having first class functions or recursion or even keyword parameters. This may be an impossible dream. These things don't get discovered that often. But I am always looking.  
  
  
  
  
  
**1\. You Can Use Whatever Language You Want.**  
  
Writing application programs used to mean writing desktop software. And in desktop software there is a big bias toward writing the application in the same language as the operating system. And so ten years ago, writing software pretty much meant writing software in C. Eventually a tradition evolved: application programs must not be written in unusual languages. And this tradition had so long to develop that nontechnical people like managers and venture capitalists also learned it.  
  
Server-based software blows away this whole model. With server-based software you can use any language you want. Almost nobody understands this yet (especially not managers and venture capitalists). A few hackers understand it, and that's why we even hear about new, indy languages like Perl and Python. We're not hearing about Perl and Python because people are using them to write Windows apps.  
  
What this means for us, as people interested in designing programming languages, is that there is now potentially an actual audience for our work.  
  
**2\. Speed Comes from Profilers.**  
  
Language designers, or at least language implementors, like to write compilers that generate fast code. But I don't think this is what makes languages fast for users. Knuth pointed out long ago that speed only matters in a few critical bottlenecks. And anyone who's tried it knows that you can't guess where these bottlenecks are. Profilers are the answer.  
  
Language designers are solving the wrong problem. Users don't need benchmarks to run fast. What they need is a language that can show them what parts of their own programs need to be rewritten. That's where speed comes from in practice. So maybe it would be a net win if language implementors took half the time they would have spent doing compiler optimizations and spent it writing a good profiler instead.  
  
**3\. You Need an Application to Drive the Design of a Language.**  
  
This may not be an absolute rule, but it seems like the best languages all evolved together with some application they were being used to write. C was written by people who needed it for systems programming. Lisp was developed partly to do symbolic differentiation, and McCarthy was so eager to get started that he was writing differentiation programs even in the first paper on Lisp, in 1960.  
  
It's especially good if your application solves some new problem. That will tend to drive your language to have new features that programmers need. I personally am interested in writing a language that will be good for writing server-based applications.  
  
\[During the panel, Guy Steele also made this point, with the additional suggestion that the application should not consist of writing the compiler for your language, unless your language happens to be intended for writing compilers.\]  
  
**4\. A Language Has to Be Good for Writing Throwaway Programs.**  
  
You know what a throwaway program is: something you write quickly for some limited task. I think if you looked around you'd find that a lot of big, serious programs started as throwaway programs. I would not be surprised if _most_ programs started as throwaway programs. And so if you want to make a language that's good for writing software in general, it has to be good for writing throwaway programs, because that is the larval stage of most software.  
  
**5\. Syntax Is Connected to Semantics.**  
  
It's traditional to think of syntax and semantics as being completely separate. This will sound shocking, but it may be that they aren't. I think that what you want in your language may be related to how you express it.  
  
I was talking recently to Robert Morris, and he pointed out that operator overloading is a bigger win in languages with infix syntax. In a language with prefix syntax, any function you define is effectively an operator. If you want to define a plus for a new type of number you've made up, you can just define a new function to add them. If you do that in a language with infix syntax, there's a big difference in appearance between the use of an overloaded operator and a function call.  
  
  
  
  
  
**1\. New Programming Languages.**  
  
Back in the 1970s it was fashionable to design new programming languages. Recently it hasn't been. But I think server-based software will make new languages fashionable again. With server-based software, you can use any language you want, so if someone does design a language that actually seems better than others that are available, there will be people who take a risk and use it.  
  
**2\. Time-Sharing.**  
  
Richard Kelsey gave this as an idea whose time has come again in the last panel, and I completely agree with him. My guess (and Microsoft's guess, it seems) is that much computing will move from the desktop onto remote servers. In other words, time-sharing is back. And I think there will need to be support for it at the language level. For example, I know that Richard and Jonathan Rees have done a lot of work implementing process scheduling within Scheme 48.  
  
**3\. Efficiency.**  
  
Recently it was starting to seem that computers were finally fast enough. More and more we were starting to hear about byte code, which implies to me at least that we feel we have cycles to spare. But I don't think we will, with server-based software. Someone is going to have to pay for the servers that the software runs on, and the number of users they can support per machine will be the divisor of their capital cost.  
  
So I think efficiency will matter, at least in computational bottlenecks. It will be especially important to do i/o fast, because server-based applications do a lot of i/o.  
  
It may turn out that byte code is not a win, in the end. Sun and Microsoft seem to be facing off in a kind of a battle of the byte codes at the moment. But they're doing it because byte code is a convenient place to insert themselves into the process, not because byte code is in itself a good idea. It may turn out that this whole battleground gets bypassed. That would be kind of amusing.  
  
  
  
  
  
**1\. Clients.**  
  
This is just a guess, but my guess is that the winning model for most applications will be purely server-based. Designing software that works on the assumption that everyone will have your client is like designing a society on the assumption that everyone will just be honest. It would certainly be convenient, but you have to assume it will never happen.  
  
I think there will be a proliferation of devices that have some kind of Web access, and all you'll be able to assume about them is that they can support simple html and forms. Will you have a browser on your cell phone? Will there be a phone in your palm pilot? Will your blackberry get a bigger screen? Will you be able to browse the Web on your gameboy? Your watch? I don't know. And I don't have to know if I bet on everything just being on the server. It's just so much more robust to have all the [brains on the server](road.html).  
  
**2\. Object-Oriented Programming.**  
  
I realize this is a controversial one, but I don't think object-oriented programming is such a big deal. I think it is a fine model for certain kinds of applications that need that specific kind of data structure, like window systems, simulations, and cad programs. But I don't see why it ought to be the model for all programming.  
  
I think part of the reason people in big companies like object-oriented programming is because it yields a lot of what looks like work. Something that might naturally be represented as, say, a list of integers, can now be represented as a class with all kinds of scaffolding and hustle and bustle.  
  
Another attraction of object-oriented programming is that methods give you some of the effect of first class functions. But this is old news to Lisp programmers. When you have actual first class functions, you can just use them in whatever way is appropriate to the task at hand, instead of forcing everything into a mold of classes and methods.  
  
What this means for language design, I think, is that you shouldn't build object-oriented programming in too deeply. Maybe the answer is to offer more general, underlying stuff, and let people design whatever object systems they want as libraries.  
  
**3\. Design by Committee.**  
  
Having your language designed by a committee is a big pitfall, and not just for the reasons everyone knows about. Everyone knows that committees tend to yield lumpy, inconsistent designs. But I think a greater danger is that they won't take risks. When one person is in charge he can take risks that a committee would never agree on.  
  
Is it necessary to take risks to design a good language though? Many people might suspect that language design is something where you should stick fairly close to the conventional wisdom. I bet this isn't true. In everything else people do, reward is proportionate to risk. Why should language design be any different?  
  

  

[Japanese Translation](http://d.hatena.ne.jp/lionfan/20070215)  
  

  

  
  

* * *



# The Roots of Lisp

[](index.html)  
  

[](https://sep.yimg.com/ca/I/paulgraham_2202_11667583)  
  
May 2001

_(I wrote this article to help myself understand exactly what McCarthy discovered. You don't need to know this stuff to program in Lisp, but it should be helpful to anyone who wants to understand the essence of Lisp � both in the sense of its origins and its semantic core. The fact that it has such a core is one of Lisp's distinguishing features, and the reason why, unlike other languages, Lisp has dialects.)_  
  
In 1960, [John McCarthy](http://www-formal.stanford.edu/jmc/index.html) published a remarkable paper in which he did for programming something like what Euclid did for geometry. He showed how, given a handful of simple operators and a notation for functions, you can build a whole programming language. He called this language Lisp, for "List Processing," because one of his key ideas was to use a simple data structure called a _list_ for both code and data.  
  
It's worth understanding what McCarthy discovered, not just as a landmark in the history of computers, but as a model for what programming is tending to become in our own time. It seems to me that there have been two really clean, consistent models of programming so far: the C model and the Lisp model. These two seem points of high ground, with swampy lowlands between them. As computers have grown more powerful, the new languages being developed have been [moving steadily](diff.html) toward the Lisp model. A popular recipe for new programming languages in the past 20 years has been to take the C model of computing and add to it, piecemeal, parts taken from the Lisp model, like runtime typing and garbage collection.  
  
In this article I'm going to try to explain in the simplest possible terms what McCarthy discovered. The point is not just to learn about an interesting theoretical result someone figured out forty years ago, but to show where languages are heading. The unusual thing about Lisp � in fact, the defining quality of Lisp � is that it can be written in itself. To understand what McCarthy meant by this, we're going to retrace his steps, with his mathematical notation translated into running Common Lisp code.  
  
  

  

[Complete Article (Postscript)](https://sep.yimg.com/ty/cdn/paulgraham/jmc.ps?t=1595850613&)  

[What Made Lisp Different](diff.html)  

[The Code](https://sep.yimg.com/ty/cdn/paulgraham/jmc.lisp?t=1595850613&)  

[Chinese Translation](http://daiyuwen.freeshell.org/gb/rol/roots_of_lisp.html)  

[Japanese Translation](http://d.hatena.ne.jp/lionfan/20070202)  

[Portuguese Translation](http://www.ciul.ul.pt/~tca/pdf/rootsoflisp.pdf)  

[Korean Translation](http://blog.java2game.com/270)  

  

  
  

* * *



# The Other Road Ahead

[](index.html)  
  

  
  
September 2001  
  
_(This article explains why much of the next generation of software may be server-based, what that will mean for programmers, and why this new kind of software is a great opportunity for startups. It's derived from a talk at BBN Labs.)_  
  
In the summer of 1995, my friend Robert Morris and I decided to start a startup. The PR campaign leading up to Netscape's IPO was running full blast then, and there was a lot of talk in the press about online commerce. At the time there might have been thirty actual stores on the Web, all made by hand. If there were going to be a lot of online stores, there would need to be software for making them, so we decided to write some.  
  
For the first week or so we intended to make this an ordinary desktop application. Then one day we had the idea of making the software run on our Web server, using the browser as an interface. We tried rewriting the software to work over the Web, and it was clear that this was the way to go. If we wrote our software to run on the server, it would be a lot easier for the users and for us as well.  
  
This turned out to be a good plan. Now, as [Yahoo Store](http://store.yahoo.com), this software is the most popular online store builder, with about 14,000 users.  
  
When we started Viaweb, hardly anyone understood what we meant when we said that the software ran on the server. It was not until Hotmail was launched a year later that people started to get it. Now everyone knows that this is a valid approach. There is a name now for what we were: an Application Service Provider, or ASP.  
  
I think that a lot of the next generation of software will be written on this model. Even Microsoft, who have the most to lose, seem to see the inevitablity of moving some things off the desktop. If software moves off the desktop and onto servers, it will mean a very different world for developers. This article describes the surprising things we saw, as some of the first visitors to this new world. To the extent software does move onto servers, what I'm describing here is the future.  
  
**The Next Thing?**  
  
When we look back on the desktop software era, I think we'll marvel at the inconveniences people put up with, just as we marvel now at what early car owners put up with. For the first twenty or thirty years, you had to be a car expert to own a car. But cars were such a big win that lots of people who weren't car experts wanted to have them as well.  
  
Computers are in this phase now. When you own a desktop computer, you end up learning a lot more than you wanted to know about what's happening inside it. But more than half the households in the US own one. My mother has a computer that she uses for email and for keeping accounts. About a year ago she was alarmed to receive a letter from Apple, offering her a discount on a new version of the operating system. There's something wrong when a sixty-five year old woman who wants to use a computer for email and accounts has to think about installing new operating systems. Ordinary users shouldn't even know the words "operating system," much less "device driver" or "patch."  
  
There is now another way to deliver software that will save users from becoming system administrators. Web-based applications are programs that run on Web servers and use Web pages as the user interface. For the average user this new kind of software will be easier, cheaper, more mobile, more reliable, and often more powerful than desktop software.  
  
With Web-based software, most users won't have to think about anything except the applications they use. All the messy, changing stuff will be sitting on a server somewhere, maintained by the kind of people who are good at that kind of thing. And so you won't ordinarily need a computer, per se, to use software. All you'll need will be something with a keyboard, a screen, and a Web browser. Maybe it will have wireless Internet access. Maybe it will also be your cell phone. Whatever it is, it will be consumer electronics: something that costs about $200, and that people choose mostly based on how the case looks. You'll pay more for Internet services than you do for the hardware, just as you do now with telephones. \[1\]  
  
It will take about a tenth of a second for a click to get to the server and back, so users of heavily interactive software, like Photoshop, will still want to have the computations happening on the desktop. But if you look at the kind of things most people use computers for, a tenth of a second latency would not be a problem. My mother doesn't really need a desktop computer, and there are a lot of people like her.  
  
**The Win for Users**  
  
Near my house there is a car with a bumper sticker that reads "death before inconvenience." Most people, most of the time, will take whatever choice requires least work. If Web-based software wins, it will be because it's more convenient. And it looks as if it will be, for users and developers both.  
  
To use a purely Web-based application, all you need is a browser connected to the Internet. So you can use a Web-based application anywhere. When you install software on your desktop computer, you can only use it on that computer. Worse still, your files are trapped on that computer. The inconvenience of this model becomes more and more evident as people get used to networks.  
  
The thin end of the wedge here was Web-based email. Millions of people now realize that you should have access to email messages no matter where you are. And if you can see your email, why not your calendar? If you can discuss a document with your colleagues, why can't you edit it? Why should any of your data be trapped on some computer sitting on a faraway desk?  
  
The whole idea of "your computer" is going away, and being replaced with "your data." You should be able to get at your data from any computer. Or rather, any client, and a client doesn't have to be a computer.  
  
Clients shouldn't store data; they should be like telephones. In fact they may become telephones, or vice versa. And as clients get smaller, you have another reason not to keep your data on them: something you carry around with you can be lost or stolen. Leaving your PDA in a taxi is like a disk crash, except that your data is handed to [someone else](http://news.zdnet.co.uk/business/0,39020645,2077931,00.htm) instead of being vaporized.  
  
With purely Web-based software, neither your data nor the applications are kept on the client. So you don't have to install anything to use it. And when there's no installation, you don't have to worry about installation going wrong. There can't be incompatibilities between the application and your operating system, because the software doesn't run on your operating system.  
  
Because it needs no installation, it will be easy, and common, to try Web-based software before you "buy" it. You should expect to be able to test-drive any Web-based application for free, just by going to the site where it's offered. At Viaweb our whole site was like a big arrow pointing users to the test drive.  
  
After trying the demo, signing up for the service should require nothing more than filling out a brief form (the briefer the better). And that should be the last work the user has to do. With Web-based software, you should get new releases without paying extra, or doing any work, or possibly even knowing about it.  
  
Upgrades won't be the big shocks they are now. Over time applications will quietly grow more powerful. This will take some effort on the part of the developers. They will have to design software so that it can be updated without confusing the users. That's a new problem, but there are ways to solve it.  
  
With Web-based applications, everyone uses the same version, and bugs can be fixed as soon as they're discovered. So Web-based software should have far fewer bugs than desktop software. At Viaweb, I doubt we ever had ten known bugs at any one time. That's orders of magnitude better than desktop software.  
  
Web-based applications can be used by several people at the same time. This is an obvious win for collaborative applications, but I bet users will start to want this in most applications once they realize it's possible. It will often be useful to let two people edit the same document, for example. Viaweb let multiple users edit a site simultaneously, more because that was the right way to write the software than because we expected users to want to, but it turned out that many did.  
  
When you use a Web-based application, your data will be safer. Disk crashes won't be a thing of the past, but users won't hear about them anymore. They'll happen within server farms. And companies offering Web-based applications will actually do backups-- not only because they'll have real system administrators worrying about such things, but because an ASP that does lose people's data will be in big, big trouble. When people lose their own data in a disk crash, they can't get that mad, because they only have themselves to be mad at. When a company loses their data for them, they'll get a lot madder.  
  
Finally, Web-based software should be less vulnerable to viruses. If the client doesn't run anything except a browser, there's less chance of running viruses, and no data locally to damage. And a program that attacked the servers themselves should find them very well defended. \[2\]  
  
For users, Web-based software will be _less stressful._ I think if you looked inside the average Windows user you'd find a huge and pretty much untapped desire for software meeting that description. Unleashed, it could be a powerful force.  
  
**City of Code**  
  
To developers, the most conspicuous difference between Web-based and desktop software is that a Web-based application is not a single piece of code. It will be a collection of programs of different types rather than a single big binary. And so designing Web-based software is like desiging a city rather than a building: as well as buildings you need roads, street signs, utilities, police and fire departments, and plans for both growth and various kinds of disasters.  
  
At Viaweb, software included fairly big applications that users talked to directly, programs that those programs used, programs that ran constantly in the background looking for problems, programs that tried to restart things if they broke, programs that ran occasionally to compile statistics or build indexes for searches, programs we ran explicitly to garbage-collect resources or to move or restore data, programs that pretended to be users (to measure performance or expose bugs), programs for diagnosing network troubles, programs for doing backups, interfaces to outside services, software that drove an impressive collection of dials displaying real-time server statistics (a hit with visitors, but indispensable for us too), modifications (including bug fixes) to open-source software, and a great many configuration files and settings. Trevor Blackwell wrote a spectacular program for moving stores to new servers across the country, without shutting them down, after we were bought by Yahoo. Programs paged us, sent faxes and email to users, conducted transactions with credit card processors, and talked to one another through sockets, pipes, http requests, ssh, udp packets, shared memory, and files. Some of Viaweb even consisted of the absence of programs, since one of the keys to Unix security is not to run unnecessary utilities that people might use to break into your servers.  
  
It did not end with software. We spent a lot of time thinking about server configurations. We built the servers ourselves, from components-- partly to save money, and partly to get exactly what we wanted. We had to think about whether our upstream ISP had fast enough connections to all the backbones. We serially [dated](http://groups.google.com/groups?selm=6hdipo%243o0%241%40FreeBSD.csie.NCTU.edu.tw) RAID suppliers.  
  
But hardware is not just something to worry about. When you control it you can do more for users. With a desktop application, you can specify certain minimum hardware, but you can't add more. If you administer the servers, you can in one step enable all your users to page people, or send faxes, or send commands by phone, or process credit cards, etc, just by installing the relevant hardware. We always looked for new ways to add features with hardware, not just because it pleased users, but also as a way to distinguish ourselves from competitors who (either because they sold desktop software, or resold Web-based applications through ISPs) didn't have direct control over the hardware.  
  
Because the software in a Web-based application will be a collection of programs rather than a single binary, it can be written in any number of different languages. When you're writing desktop software, you're practically forced to write the application in the same language as the underlying operating system-- meaning C and C++. And so these languages (especially among nontechnical people like managers and VCs) got to be considered as the languages for "serious" software development. But that was just an artifact of the way desktop software had to be delivered. For server-based software you can use any language you want. \[3\] Today a lot of the top hackers are using languages far removed from C and C++: Perl, Python, and even Lisp.  
  
With server-based software, no one can tell you what language to use, because you control the whole system, right down to the hardware. Different languages are good for different tasks. You can use whichever is best for each. And when you have competitors, "you can" means "you must" (we'll return to this later), because if you don't take advantage of this possibility, your competitors will.  
  
Most of our competitors used C and C++, and this made their software visibly inferior because (among other things), they had no way around the statelessness of CGI scripts. If you were going to change something, all the changes had to happen on one page, with an Update button at the bottom. As I've written elsewhere, by using [Lisp](avg.html), which many people still consider a research language, we could make the Viaweb editor behave more like desktop software.  
  
**Releases**  
  
One of the most important changes in this new world is the way you do releases. In the desktop software business, doing a release is a huge trauma, in which the whole company sweats and strains to push out a single, giant piece of code. Obvious comparisons suggest themselves, both to the process and the resulting product.  
  
With server-based software, you can make changes almost as you would in a program you were writing for yourself. You release software as a series of incremental changes instead of an occasional big explosion. A typical desktop software company might do one or two releases a year. At Viaweb we often did three to five releases a day.  
  
When you switch to this new model, you realize how much software development is affected by the way it is released. Many of the nastiest problems you see in the desktop software business are due to catastrophic nature of releases.  
  
When you release only one new version a year, you tend to deal with bugs wholesale. Some time before the release date you assemble a new version in which half the code has been torn out and replaced, introducing countless bugs. Then a squad of QA people step in and start counting them, and the programmers work down the list, fixing them. They do not generally get to the end of the list, and indeed, no one is sure where the end is. It's like fishing rubble out of a pond. You never really know what's happening inside the software. At best you end up with a statistical sort of correctness.  
  
With server-based software, most of the change is small and incremental. That in itself is less likely to introduce bugs. It also means you know what to test most carefully when you're about to release software: the last thing you changed. You end up with a much firmer grip on the code. As a general rule, you do know what's happening inside it. You don't have the source code memorized, of course, but when you read the source you do it like a pilot scanning the instrument panel, not like a detective trying to unravel some mystery.  
  
Desktop software breeds a certain fatalism about bugs. You know that you're shipping something loaded with bugs, and you've even set up mechanisms to compensate for it (e.g. patch releases). So why worry about a few more? Soon you're releasing whole features you know are broken. [Apple](http://news.cnet.com/news/0-1006-200-5195914.html) did this earlier this year. They felt under pressure to release their new OS, whose release date had already slipped four times, but some of the software (support for CDs and DVDs) wasn't ready. The solution? They released the OS without the unfinished parts, and users will have to install them later.  
  
With Web-based software, you never have to release software before it works, and you can release it as soon as it does work.  
  
The industry veteran may be thinking, it's a fine-sounding idea to say that you never have to release software before it works, but what happens when you've promised to deliver a new version of your software by a certain date? With Web-based software, you wouldn't make such a promise, because there are no versions. Your software changes gradually and continuously. Some changes might be bigger than others, but the idea of versions just doesn't naturally fit onto Web-based software.  
  
If anyone remembers Viaweb this might sound odd, because we were always announcing new versions. This was done entirely for PR purposes. The trade press, we learned, thinks in version numbers. They will give you major coverage for a major release, meaning a new first digit on the version number, and generally a paragraph at most for a point release, meaning a new digit after the decimal point.  
  
Some of our competitors were offering desktop software and actually had version numbers. And for these releases, the mere fact of which seemed to us evidence of their backwardness, they would get all kinds of publicity. We didn't want to miss out, so we started giving version numbers to our software too. When we wanted some publicity, we'd make a list of all the features we'd added since the last "release," stick a new version number on the software, and issue a press release saying that the new version was available immediately. Amazingly, no one ever called us on it.  
  
By the time we were bought, we had done this three times, so we were on Version 4. Version 4.1 if I remember correctly. After Viaweb became Yahoo Store, there was no longer such a desperate need for publicity, so although the software continued to evolve, the whole idea of version numbers was quietly dropped.  
  
**Bugs**  
  
The other major technical advantage of Web-based software is that you can reproduce most bugs. You have the users' data right there on your disk. If someone breaks your software, you don't have to try to guess what's going on, as you would with desktop software: you should be able to reproduce the error while they're on the phone with you. You might even know about it already, if you have code for noticing errors built into your application.  
  
Web-based software gets used round the clock, so everything you do is immediately put through the wringer. Bugs turn up quickly.  
  
Software companies are sometimes accused of letting the users debug their software. And that is just what I'm advocating. For Web-based software it's actually a good plan, because the bugs are fewer and transient. When you release software gradually you get far fewer bugs to start with. And when you can reproduce errors and release changes instantly, you can find and fix most bugs as soon as they appear. We never had enough bugs at any one time to bother with a formal bug-tracking system.  
  
You should test changes before you release them, of course, so no major bugs should get released. Those few that inevitably slip through will involve borderline cases and will only affect the few users that encounter them before someone calls in to complain. As long as you fix bugs right away, the net effect, for the average user, is far fewer bugs. I doubt the average Viaweb user ever saw a bug.  
  
Fixing fresh bugs is easier than fixing old ones. It's usually fairly quick to find a bug in code you just wrote. When it turns up you often know what's wrong before you even look at the source, because you were already worrying about it subconsciously. Fixing a bug in something you wrote six months ago (the average case if you release once a year) is a lot more work. And since you don't understand the code as well, you're more likely to fix it in an ugly way, or even introduce more bugs. \[4\]  
  
When you catch bugs early, you also get fewer compound bugs. Compound bugs are two separate bugs that interact: you trip going downstairs, and when you reach for the handrail it comes off in your hand. In software this kind of bug is the hardest to find, and also tends to have the worst consequences. \[5\] The traditional "break everything and then filter out the bugs" approach inherently yields a lot of compound bugs. And software that's released in a series of small changes inherently tends not to. The floors are constantly being swept clean of any loose objects that might later get stuck in something.  
  
It helps if you use a technique called functional programming. Functional programming means avoiding side-effects. It's something you're more likely to see in research papers than commercial software, but for Web-based applications it turns out to be really useful. It's hard to write entire programs as purely functional code, but you can write substantial chunks this way. It makes those parts of your software easier to test, because they have no state, and that is very convenient in a situation where you are constantly making and testing small modifications. I wrote much of Viaweb's editor in this style, and we made our scripting language, [RTML](http://store.yahoo.com/rtml.html), a purely functional language.  
  
People from the desktop software business will find this hard to credit, but at Viaweb bugs became almost a game. Since most released bugs involved borderline cases, the users who encountered them were likely to be advanced users, pushing the envelope. Advanced users are more forgiving about bugs, especially since you probably introduced them in the course of adding some feature they were asking for. In fact, because bugs were rare and you had to be doing sophisticated things to see them, advanced users were often proud to catch one. They would call support in a spirit more of triumph than anger, as if they had scored points off us.  
  
**Support**  
  
When you can reproduce errors, it changes your approach to customer support. At most software companies, support is offered as a way to make customers feel better. They're either calling you about a known bug, or they're just doing something wrong and you have to figure out what. In either case there's not much you can learn from them. And so you tend to view support calls as a pain in the ass that you want to isolate from your developers as much as possible.  
  
This was not how things worked at Viaweb. At Viaweb, support was free, because we wanted to hear from customers. If someone had a problem, we wanted to know about it right away so that we could reproduce the error and release a fix.  
  
So at Viaweb the developers were always in close contact with support. The customer support people were about thirty feet away from the programmers, and knew that they could always interrupt anything with a report of a genuine bug. We would leave a board meeting to fix a serious bug.  
  
Our approach to support made everyone happier. The customers were delighted. Just imagine how it would feel to call a support line and be treated as someone bringing important news. The customer support people liked it because it meant they could help the users, instead of reading scripts to them. And the programmers liked it because they could reproduce bugs instead of just hearing vague second-hand reports about them.  
  
Our policy of fixing bugs on the fly changed the relationship between customer support people and hackers. At most software companies, support people are underpaid human shields, and hackers are little copies of God the Father, creators of the world. Whatever the procedure for reporting bugs, it is likely to be one-directional: support people who hear about bugs fill out some form that eventually gets passed on (possibly via QA) to programmers, who put it on their list of things to do. It was very different at Viaweb. Within a minute of hearing about a bug from a customer, the support people could be standing next to a programmer hearing him say "Shit, you're right, it's a bug." It delighted the support people to hear that "you're right" from the hackers. They used to bring us bugs with the same expectant air as a cat bringing you a mouse it has just killed. It also made them more careful in judging the seriousness of a bug, because now their honor was on the line.  
  
After we were bought by Yahoo, the customer support people were moved far away from the programmers. It was only then that we realized that they were effectively QA and to some extent marketing as well. In addition to catching bugs, they were the keepers of the knowledge of vaguer, buglike things, like features that confused users. \[6\] They were also a kind of proxy focus group; we could ask them which of two new features users wanted more, and they were always right.  
  
**Morale**  
  
Being able to release software immediately is a big motivator. Often as I was walking to work I would think of some change I wanted to make to the software, and do it that day. This worked for bigger features as well. Even if something was going to take two weeks to write (few projects took longer), I knew I could see the effect in the software as soon as it was done.  
  
If I'd had to wait a year for the next release, I would have shelved most of these ideas, for a while at least. The thing about ideas, though, is that they lead to more ideas. Have you ever noticed that when you sit down to write something, half the ideas that end up in it are ones you thought of while writing it? The same thing happens with software. Working to implement one idea gives you more ideas. So shelving an idea costs you not only that delay in implementing it, but also all the ideas that implementing it would have led to. In fact, shelving an idea probably even inhibits new ideas: as you start to think of some new feature, you catch sight of the shelf and think "but I already have a lot of new things I want to do for the next release."  
  
What big companies do instead of implementing features is plan them. At Viaweb we sometimes ran into trouble on this account. Investors and analysts would ask us what we had planned for the future. The truthful answer would have been, we didn't have any plans. We had general ideas about things we wanted to improve, but if we knew how we would have done it already. What were we going to do in the next six months? Whatever looked like the biggest win. I don't know if I ever dared give this answer, but that was the truth. Plans are just another word for ideas on the shelf. When we thought of good ideas, we implemented them.  
  
At Viaweb, as at many software companies, most code had one definite owner. But when you owned something you really owned it: no one except the owner of a piece of software had to approve (or even know about) a release. There was no protection against breakage except the fear of looking like an idiot to one's peers, and that was more than enough. I may have given the impression that we just blithely plowed forward writing code. We did go fast, but we thought very carefully before we released software onto those servers. And paying attention is more important to reliability than moving slowly. Because he pays close attention, a Navy pilot can land a 40,000 lb. aircraft at 140 miles per hour on a pitching carrier deck, at night, more safely than the average teenager can cut a bagel.  
  
This way of writing software is a double-edged sword of course. It works a lot better for a small team of good, trusted programmers than it would for a big company of mediocre ones, where bad ideas are caught by committees instead of the people that had them.  
  
**Brooks in Reverse**  
  
Fortunately, Web-based software does require fewer programmers. I once worked for a medium-sized desktop software company that had over 100 people working in engineering as a whole. Only 13 of these were in product development. All the rest were working on releases, ports, and so on. With Web-based software, all you need (at most) are the 13 people, because there are no releases, ports, and so on.  
  
Viaweb was written by just three people. \[7\] I was always under pressure to hire more, because we wanted to get bought, and we knew that buyers would have a hard time paying a high price for a company with only three programmers. (Solution: we hired more, but created new projects for them.)  
  
When you can write software with fewer programmers, it saves you more than money. As Fred Brooks pointed out in _The Mythical Man-Month,_ adding people to a project tends to slow it down. The number of possible connections between developers grows exponentially with the size of the group. The larger the group, the more time they'll spend in meetings negotiating how their software will work together, and the more bugs they'll get from unforeseen interactions. Fortunately, this process also works in reverse: as groups get smaller, software development gets exponentially more efficient. I can't remember the programmers at Viaweb ever having an actual meeting. We never had more to say at any one time than we could say as we were walking to lunch.  
  
If there is a downside here, it is that all the programmers have to be to some degree system administrators as well. When you're hosting software, someone has to be watching the servers, and in practice the only people who can do this properly are the ones who wrote the software. At Viaweb our system had so many components and changed so frequently that there was no definite border between software and infrastructure. Arbitrarily declaring such a border would have constrained our design choices. And so although we were constantly hoping that one day ("in a couple months") everything would be stable enough that we could hire someone whose job was just to worry about the servers, it never happened.  
  
I don't think it could be any other way, as long as you're still actively developing the product. Web-based software is never going to be something you write, check in, and go home. It's a live thing, running on your servers right now. A bad bug might not just crash one user's process; it could crash them all. If a bug in your code corrupts some data on disk, you have to fix it. And so on. We found that you don't have to watch the servers every minute (after the first year or so), but you definitely want to keep an eye on things you've changed recently. You don't release code late at night and then go home.  
  
**Watching Users**  
  
With server-based software, you're in closer touch with your code. You can also be in closer touch with your users. Intuit is famous for introducing themselves to customers at retail stores and asking to follow them home. If you've ever watched someone use your software for the first time, you know what surprises must