6 ms·

Why Lisp Is Unpopular

62 points by mnemonicsloth 19y ago

A picture is worth 1K - 3 * 2^3 words: http://www.hq.usace.army.mil/cepa/pubs/jul00/radar.jpg That's a phased-array radar installation that (I understand) was built to detect Soviet ballistic missiles inbound over the North Pole. Google has more images. The term "Phased Array" has interesting implications here. It means that received phase information is preserved among the many small antenna elements on each face of the structure. That only works if the paths traversed by the antenna outputs are all the same length (modulo lambda), with a margin of error around .1 * lambda, where lambda is a representative wavelength of the received signal. This thing uses microwave signals [1], which means we can reformulate the preceding paragraph as follows: You're looking at a picture of a single circuit the size of a large office building, built to a precision of 100 microns. Now's a good time to point out that little yellow thing in the lower left corner of the image is probably some kind of earth-moving machine. The US military built a number of these installations in Alaska and Canada during the Cold War. Let's pause here to think about all the structure-hardening and weatherproofing that must have been involved. What does this have to do with Lisp? It's not an exaggeration to say that building and maintaining these early warning systems required the attention of at least a third of all American microwave engineers [2]. When the Cold War ended, most of them left the big contractors the DoD hired to build the radar and started doing basically the same work for cellular companies [3]. Ten or fifteen years later, mobile phones became ubiquitous. In the case of Lisp, there was no continuity in the transition. Common Lisp, in particular, was primarily in use by a rarefied group of specialists working on room-sized computers at places like DARPA, thinking about things like AI for driving tanks across central Germany. When the funding dried up and the specialists had to move on [4], they found work writing C, Perl, or Java on microcomputers. So Lisp lost its user base and the last of its major hardware platforms (now that you couldn't buy a Lisp Machine any more) all at once. Of course, when programmers discuss Lisp's continuing lack of popularity, regardless of their opinion of the language itself, they seem a lot more willing to blame things programmers ultimately control, like the social habits of Lisp users or the damn parentheses. [1] I'm not a microwave guy, so some of this explanation is simplistic almost to the point of inaccuracy. In particular, I don't know what frequency bands these installations use. This is as good a place as any to point out that I also have no special knowledge about Lisp at big organizations after the end of the Cold War. People who do are strongly encouraged to email me if I've gotten anything wrong. [2] I don't have a citation for this number, but one of my old professors does. I'll see if I can get him to email it to me. [3] "3G" cellular technology would be unworkable without big antenna arrays [4] A lot of the "classic" books on Lisp and related topics were written during this period. SICP, PAIP, and On Lisp, and The Seasoned Schemer come to mind.

pg 19y ago
I don't understand. Is the claim that Lisp is less popular now because years ago large numbers of programmers working on government-sponsored projects had to use it, but now proportionately fewer programmers work on government-sponsored projects?
- wanorris 19y ago
  Yes, I believe the claim is that when huge government projects like this were completed, all the Lisp programmers left, and had to retrain to find non-Lisp jobs. This sounds similar to the "AI collapsed and took Lisp with it" argument.
  - pg 19y ago
    There may be something to that argument, but another probably bigger factor is that there are now other dynamic languages. In 1990 there was no Python or Ruby. If you wanted to program in that kind of high level style, the only options were Lisp and Smalltalk.
    - jmatt 19y ago
      All the programmers I know who wrote lisp in the 80s, now own software companies or are in management. That may explain some of it.
    - mnemonicsloth 19y ago
      The problem with the AI winter argument is that it's half an answer that satisfies you enough to stop asking questions. If you invoke the collapse of commercial AI to explain Lisp's current state of disfavor, you then have to explain why the collapse of the dotcom bubble didn't have a similar effect on Perl or Java. Anthropomorphically, you could say that Perl retreated to its base of support among sysadmins, (Common) Lisp took shelter in big-think research organizations, and Java reinvented itself as New Cobol. But that doesn't really satisfy either, because it doesn't tell us why nobody tried to similarly re-purpose Lisp. That's especially odd, in fact, because now Everybody Knows that a lisp core is really simple, and Lisp was still on the radar in academia, where people can be rewarded for re-implementing old ideas. The problem with the AI winter argument is that it's half an answer that satisfies you enough to stop asking questions. If you invoke the collapse of commercial AI to explain Lisp's current state of disfavor, you then have to explain why the collapse of the dotcom bubble didn't have a similar effect on Perl or Java. Anthropomorphically, you could say that Perl retreated to its base of support among sysadmins, (Common) Lisp took shelter in big-think research organizations, and Java reinvented itself as New Cobol. But that doesn't really satisfy either, because it doesn't tell us why nobody tried to similarly re-purpose Lisp when DARPA et al lost funding. That's especially odd since Everybody Knows that a lisp core is really simple, and Lisp was still on the radar in academia, where people (called grad students) can be rewarded for re-implementing old ideas. I think there is a pretty good explanation (with evidence, even!), but it's not a satisfying one in this community: the end of the Cold War caused a major reallocation of scarce resources away from expensive one-off projects for the government, and towards mass-market stuff for consumers and businesses (see below), and Lisp suffered from some initial disadvantages in the new environment: * Limited availability on PC platforms * Windows Apps were sexier. The internet was sexier. * Glut of new CS students while the bubble was inflating were mostly taught in Java. (Many of them are sensitive about arguments that lend further support to the idea that their education has turned out to be less valuable than they thought at the time. That may just be my bias showing, though. I think that Math should be relabeled CS, and that "programming" is for autodidacts and trade-schoolers). * CS is a new field, and industry tends to dumb down research in new fields for a generation or so after disruptive technical innovations. That way, they get enough of the low hanging fruit that it further disruptive innovation is unlikely and their capital is safer. See below, particularly the chapter on the relationship between GE and MIT before WWII. Social arguments are hard to make here though, because so many people are so hooked on the "weird and scary and only for elitist pricks" narrative. That's kind of ironic, because high level corporate executives, if they heard the current public stand on lisp -- "it may be really productive, but it's too foreign for most people and I don't feel like learning it" -- could probably force adoption from above, especially now that cost-cutting is the order of the day. http://www.amazon.com/Leonardo-Internet-Technology-Culture-Renaissance/dp/0801878098 http://www.amazon.com/Leonardo-Internet-Technology-Culture-R...

62 points by mnemonicsloth 19y ago

pg 19y ago

I don't understand. Is the claim that Lisp is less popular now because years ago large numbers of programmers working on government-sponsored projects had to use it, but now proportionately fewer programmers work on government-sponsored projects?

wanorris 19y ago
Yes, I believe the claim is that when huge government projects like this were completed, all the Lisp programmers left, and had to retrain to find non-Lisp jobs. This sounds similar to the "AI collapsed and took Lisp with it" argument.
- pg 19y ago
  There may be something to that argument, but another probably bigger factor is that there are now other dynamic languages. In 1990 there was no Python or Ruby. If you wanted to program in that kind of high level style, the only options were Lisp and Smalltalk.
  - jmatt 19y ago
    All the programmers I know who wrote lisp in the 80s, now own software companies or are in management. That may explain some of it.
  - mnemonicsloth 19y ago
    The problem with the AI winter argument is that it's half an answer that satisfies you enough to stop asking questions. If you invoke the collapse of commercial AI to explain Lisp's current state of disfavor, you then have to explain why the collapse of the dotcom bubble didn't have a similar effect on Perl or Java. Anthropomorphically, you could say that Perl retreated to its base of support among sysadmins, (Common) Lisp took shelter in big-think research organizations, and Java reinvented itself as New Cobol. But that doesn't really satisfy either, because it doesn't tell us why nobody tried to similarly re-purpose Lisp. That's especially odd, in fact, because now Everybody Knows that a lisp core is really simple, and Lisp was still on the radar in academia, where people can be rewarded for re-implementing old ideas. The problem with the AI winter argument is that it's half an answer that satisfies you enough to stop asking questions. If you invoke the collapse of commercial AI to explain Lisp's current state of disfavor, you then have to explain why the collapse of the dotcom bubble didn't have a similar effect on Perl or Java. Anthropomorphically, you could say that Perl retreated to its base of support among sysadmins, (Common) Lisp took shelter in big-think research organizations, and Java reinvented itself as New Cobol. But that doesn't really satisfy either, because it doesn't tell us why nobody tried to similarly re-purpose Lisp when DARPA et al lost funding. That's especially odd since Everybody Knows that a lisp core is really simple, and Lisp was still on the radar in academia, where people (called grad students) can be rewarded for re-implementing old ideas. I think there is a pretty good explanation (with evidence, even!), but it's not a satisfying one in this community: the end of the Cold War caused a major reallocation of scarce resources away from expensive one-off projects for the government, and towards mass-market stuff for consumers and businesses (see below), and Lisp suffered from some initial disadvantages in the new environment: * Limited availability on PC platforms * Windows Apps were sexier. The internet was sexier. * Glut of new CS students while the bubble was inflating were mostly taught in Java. (Many of them are sensitive about arguments that lend further support to the idea that their education has turned out to be less valuable than they thought at the time. That may just be my bias showing, though. I think that Math should be relabeled CS, and that "programming" is for autodidacts and trade-schoolers). * CS is a new field, and industry tends to dumb down research in new fields for a generation or so after disruptive technical innovations. That way, they get enough of the low hanging fruit that it further disruptive innovation is unlikely and their capital is safer. See below, particularly the chapter on the relationship between GE and MIT before WWII. Social arguments are hard to make here though, because so many people are so hooked on the "weird and scary and only for elitist pricks" narrative. That's kind of ironic, because high level corporate executives, if they heard the current public stand on lisp -- "it may be really productive, but it's too foreign for most people and I don't feel like learning it" -- could probably force adoption from above, especially now that cost-cutting is the order of the day. http://www.amazon.com/Leonardo-Internet-Technology-Culture-Renaissance/dp/0801878098 http://www.amazon.com/Leonardo-Internet-Technology-Culture-R...