4 ms·

I outlined some of the problems a while ago: http://www.codexon.com/posts/why-arent-functional-languages-like-haskell-and-ocaml-popular http://www.codexon.com/

by codexon 16y ago

I outlined some of the problems a while ago: http://www.codexon.com/posts/why-arent-functional-languages-like-haskell-and-ocaml-popular http://www.codexon.com/posts/why-arent-functional-languages-... Short version: Haskell is harder than C while being slower and uses more memory. The main implementation GHC, comes with GPL concerns.

dons 16y ago
GPL concerns? Its BSD licensed. Haskell's library system is perhaps uniquely BSD-friendly, with over 75% of libraries BSD licensed. And as pointed out at the time, commercial users already funded the -dynamic flag, along with the ability to swap out libgmp. So that's FUD. As was pointed out in this forum at the time: http://news.ycombinator.com/item?id=847120 http://news.ycombinator.com/item?id=847120
- codexon 16y ago
  The default behavior was to statically link to libgmp. I don't know about you, but making people's programs GPL by default doesn't exactly inspire confidence in the industry.
  - dons 16y ago
    Whether or not libgmp was ever statically linked by default depended on your distribution. Were you shipping apps on Windows prior to the -dynamic flag's introduction? If so, then you were watching what libraries you were using, and following these instructions: http://haskell.forkio.com/gmpwindows http://haskell.forkio.com/gmpwindows These days you can just run "cabal license-check" (IIRC) which will type check all the libraries you use for compliance. Have a nice day.
    - codexon 16y ago
      Dons, I know you are devoted to Haskell and will not take any criticism of it lightly. But I am not your enemy. I am willing to praise it myself when it is more suitable. Not when you try to spin it. You and your supporters succeeded in bombarding every article that even mentions Haskell the wrong way with snide comments, and probably won over a few hobbyists. At the end of the day though, the managers making the decisions will not think "Don Stewart said so, it must be true". They will find the same conclusions that Jon Harrop, myself, and other people you have ostracized have found. And this will continue until it even if you succeed in suppressing all negative feedback on Haskell. Have a nice day.
      - loup-vaillant 16y ago
        The only thing I ever saw Don Stewart "bombing" were factual evidence. And what he destroys by truth should be annihilated anyway (P. C. Hodgell). "Don Stewart said so, it must be true" actually isn't such a bad heuristic (as far as Haskell is concerned). From what I've seen, Don is quite cautious.
        codexon 16y ago
        The only thing I ever saw Don Stewart "bombing" were factual evidence. This is not surprising coming from someone involved with Haskell. Do I really need to point to you the bug ticket that it was only 8 months ago that you could use something other than GMP? http://hackage.haskell.org/trac/ghc/ticket/601 http://hackage.haskell.org/trac/ghc/ticket/601 Even with the option to dynamically link, it still doesn't change the fact that it was the __default__. And what a scary default that was. Company lawyers don't like to touch anything related to GPL with a 10-foot pole, for the reason that it is really up to the jury to decide what is a "derivative work" even if you only link dynamically! Maybe if you live in France, or have a 5 person company, this is an acceptable state. Anyway I am done, you and Dons win, are you happy? I simply have no ulterior motive or incentive to defend my findings against organized groups who have their livelihoods and PhDs based on Haskell.
        loup-vaillant 16y ago
        Err, I was only objecting to your attack on Don's on-line behaviour. Actually I'm (still) an outsider. Consider "what I ever saw" as anecdotal evidence. Now it's a pity that companies are scared of the GPL. As far as I know, most software is custom or private, is never released[1], and thus can't possibly infringe the GPL. I think the real problem here is corporate culture. No company would object using the glibc in a proprietary program. So why would they be scared of any other LGPL library? When they don't even release the software? That's plainly irrational. Now our choice is hard, but simple: get rid of the "GPL == we can't use it" line of thinking, or get rid of restrictive licences. [1]: Many custom software actually belong to the company that wrote it, rather than to the company (or government) that purchased it. In this case GPL infringement could happen. But really, I can't fathom why some customers still don't demand complete ownership (including source code) for their custom software. That strikes me as either wildly misinformed or incredibly silly.
jrockway 16y ago
Interestingly, I came to the exact opposite conclusion. Haskell is easier than C++ (less code and more safety), and it is equally fast. Haskell may not be the perfect language for your generic FFT library to be distributed with your OS, but it's a great choice for building applications that would otherwise be C++ or Java. Java and C++ had no trouble catching on, despite being slower than C, and Haskell isn't even that much slower than C.
- vicaya 16y ago
  You know, it'd be more convincing if you can come up with reproducible benchmarks to backup your claims. According to This guy's data: http://www.codexon.com/posts/debunking-the-erlang-and-haskell-hype-for-servers http://www.codexon.com/posts/debunking-the-erlang-and-haskel... Haskell is even slower than python (which is not known for its speed) for implementing simple servers. Haskell code that performs close (within an order of magnitude in terms of run time and space) to its C/C++ equivalent typically is also close in code length (and often less readable!) and often unsafe as well (requires explicit unboxing etc.)
  - applicative 16y ago
    If you read a little further, you will discover that there is some discussion of his 'data'.

by codexon 16y ago

dons 16y ago

GPL concerns? Its BSD licensed. Haskell's library system is perhaps uniquely BSD-friendly, with over 75% of libraries BSD licensed. And as pointed out at the time, commercial users already funded the -dynamic flag, along with the ability to swap out libgmp. So that's FUD. As was pointed out in this forum at the time: http://news.ycombinator.com/item?id=847120 http://news.ycombinator.com/item?id=847120

codexon 16y ago
The default behavior was to statically link to libgmp. I don't know about you, but making people's programs GPL by default doesn't exactly inspire confidence in the industry.
- dons 16y ago
  Whether or not libgmp was ever statically linked by default depended on your distribution. Were you shipping apps on Windows prior to the -dynamic flag's introduction? If so, then you were watching what libraries you were using, and following these instructions: http://haskell.forkio.com/gmpwindows http://haskell.forkio.com/gmpwindows These days you can just run "cabal license-check" (IIRC) which will type check all the libraries you use for compliance. Have a nice day.
  - codexon 16y ago
    Dons, I know you are devoted to Haskell and will not take any criticism of it lightly. But I am not your enemy. I am willing to praise it myself when it is more suitable. Not when you try to spin it. You and your supporters succeeded in bombarding every article that even mentions Haskell the wrong way with snide comments, and probably won over a few hobbyists. At the end of the day though, the managers making the decisions will not think "Don Stewart said so, it must be true". They will find the same conclusions that Jon Harrop, myself, and other people you have ostracized have found. And this will continue until it even if you succeed in suppressing all negative feedback on Haskell. Have a nice day.
    - loup-vaillant 16y ago
      The only thing I ever saw Don Stewart "bombing" were factual evidence. And what he destroys by truth should be annihilated anyway (P. C. Hodgell). "Don Stewart said so, it must be true" actually isn't such a bad heuristic (as far as Haskell is concerned). From what I've seen, Don is quite cautious.
      - codexon 16y ago
        The only thing I ever saw Don Stewart "bombing" were factual evidence. This is not surprising coming from someone involved with Haskell. Do I really need to point to you the bug ticket that it was only 8 months ago that you could use something other than GMP? http://hackage.haskell.org/trac/ghc/ticket/601 http://hackage.haskell.org/trac/ghc/ticket/601 Even with the option to dynamically link, it still doesn't change the fact that it was the __default__. And what a scary default that was. Company lawyers don't like to touch anything related to GPL with a 10-foot pole, for the reason that it is really up to the jury to decide what is a "derivative work" even if you only link dynamically! Maybe if you live in France, or have a 5 person company, this is an acceptable state. Anyway I am done, you and Dons win, are you happy? I simply have no ulterior motive or incentive to defend my findings against organized groups who have their livelihoods and PhDs based on Haskell.
        loup-vaillant 16y ago
        Err, I was only objecting to your attack on Don's on-line behaviour. Actually I'm (still) an outsider. Consider "what I ever saw" as anecdotal evidence. Now it's a pity that companies are scared of the GPL. As far as I know, most software is custom or private, is never released[1], and thus can't possibly infringe the GPL. I think the real problem here is corporate culture. No company would object using the glibc in a proprietary program. So why would they be scared of any other LGPL library? When they don't even release the software? That's plainly irrational. Now our choice is hard, but simple: get rid of the "GPL == we can't use it" line of thinking, or get rid of restrictive licences. [1]: Many custom software actually belong to the company that wrote it, rather than to the company (or government) that purchased it. In this case GPL infringement could happen. But really, I can't fathom why some customers still don't demand complete ownership (including source code) for their custom software. That strikes me as either wildly misinformed or incredibly silly.

jrockway 16y ago

Interestingly, I came to the exact opposite conclusion. Haskell is easier than C++ (less code and more safety), and it is equally fast. Haskell may not be the perfect language for your generic FFT library to be distributed with your OS, but it's a great choice for building applications that would otherwise be C++ or Java. Java and C++ had no trouble catching on, despite being slower than C, and Haskell isn't even that much slower than C.

vicaya 16y ago
You know, it'd be more convincing if you can come up with reproducible benchmarks to backup your claims. According to This guy's data: http://www.codexon.com/posts/debunking-the-erlang-and-haskell-hype-for-servers http://www.codexon.com/posts/debunking-the-erlang-and-haskel... Haskell is even slower than python (which is not known for its speed) for implementing simple servers. Haskell code that performs close (within an order of magnitude in terms of run time and space) to its C/C++ equivalent typically is also close in code length (and often less readable!) and often unsafe as well (requires explicit unboxing etc.)
- applicative 16y ago
  If you read a little further, you will discover that there is some discussion of his 'data'.