5 ms·

The LISP people should have changed those names way back when. Not 79, 59. McCarthy himself said then that this representation is independent of the IBM 704 com

The LISP people should have changed those names way back when. Not 79, 59. McCarthy himself said then that this representation is independent of the IBM 704 computer, or of any other electronic computer [1]. So why stick with the names of 704 registers? Was there a purpose to this? [1] https://courses.engr.illinois.edu/cs421/sp2012/project/mccarthy-lisp.pdf https://courses.engr.illinois.edu/cs421/sp2012/project/mccar...

lispm 9y ago
Lisp has both CAR/CDR and FIRST/REST. It's usual programming style to use CAR/CDR when working with CONS trees/graphs and FIRST/REST when working with lists.
- AnimalMuppet 9y ago
  OK, but isn't CalChris's point still valid? Something other than CAR/CDR would have been a better choice for working with CONS cells, even back in 1959. Do I think it's a big problem? Not really. Are the names less than optimal? Yes.
  - kbp 9y ago
    > Something other than CAR/CDR would have been a better choice for working with CONS cells, even back in 1959. > Are the names less than optimal? Yes. I don't think that's self-evident. The only better alternatives I've ever heard suggested are something like left and right, which I think are probably slightly clearer names for what the functions do, but not significantly so (I don't think "a cons is a thing with a left and right side" is that much more intuitive than "a cons is a thing with two parts called a car and a cdr"). Further, I think in 1959 when most everyone using Lisp was hacking on the implementation itself, names that are mnemonic for what the machine they were actually using is actually doing make sense; I don't think they're clearly sub-optimal now, but I think they were even less so in 1959.
    - AnimalMuppet 9y ago
      I assert that naming the interface for the implementation is almost always the wrong thing to do. (And then you ask me what would be a better name, and I don't have a perfect answer. "left" and "right" are the best I know of...)
      - kbp 9y ago
        > I assert that naming the interface for the implementation is almost always the wrong thing to do. I think it's perfectly acceptable if there aren't really any better names for the abstraction you're creating. car and cdr work just fine as made-up arbitrary names for an abstraction that doesn't really have any more natural names for them; the fact that 60 years ago they weren't actually entirely arbitrary doesn't really change that.
  - kazinator 9y ago
    These names were very well thought out. They were part of a larger vocabulary in fact; most of that vocabulary died out and only car and cdr remained. There were things like cwr, ctr and others. MacCarthy and gang were well aware (of all people!) that the language is machine-independent and not something that will be forever stuck to the particular IBM machine; they used these names anyway because there was nothing wrong with them. By the way, the C and R are not machine specific; any machine can have a "register" and any register can have "contents". On any machine where we implement cons cells, we can just call the two fields registers, understood as being A and D. These, in turn, can be "ante" and "de", if someone desperately needs mnemonics: https://en.wiktionary.org/wiki/ante https://en.wiktionary.org/wiki/ante https://www.etymonline.com/word/de- https://www.etymonline.com/word/de- "active word-forming element in English and in many words inherited from French and Latin, from Latin de "down, down from, from, off; concerning."
    - DonaldFisk 9y ago
      In addition to car and cdr, there were cpr (Contents of the Prefix part of Register) and ctr (Contents of the Tag part of Register). > By the way, the C and R are not machine specific; any machine can have a "register" Slight nitpick: the first machine on which I had a decent Lisp to use (Portable Standard Lisp) was a Burroughs 6800, a stack machine with no general purpose registers, or indeed any registers directly accessible by the programmer.
      - kazinator 9y ago
        Vi has "registers" and so do TeX and troff. I think once upon a time, it was more common to use the word for things beside CPU registers and I/O ports. If a cons cell is a context with two registers (R) whose contents we can access (C), the only vestiges of that IBM machine are the A and D letters sandwiched in between to distinguish them. Compared to how option letters change meaning between Unix commands, it's nothing. -h help? Nope, h)uman readable sizes. ^ to anchor regex at beginning; $ for end. Because on common keyboards, ^ is on the right, and $ on the left!
kbp 9y ago
> So why stick with the names of 704 registers? Was there a purpose to this? Are there other, much better names for the two parts of a pair? First and second aren't great because they're not clear that it's a pair and not a larger sequence (and the difference is important in Lisp; since lists are built on conses, it would be odd to get the tail of a list by calling 'second'). The only other names I can think of would be something like left and right, but I don't think those would be substantially easier to understand for beginners than car and cdr. When working with lists, many people prefer to use the functions first and rest, which behave identically to car and cdr but are more meaningfully named for list applications. They would be terrible names for the general cons-handling functions, though.
- lisper 9y ago
  I use FST and RST (for FIRST and REST). I like them because they are short, mnemonic, and can be composed like CAR and CDR, e.g. CADADDR = FRFRRST (though I also believe that if you find yourself doing this you're almost certainly doing something wrong).
- Steltek 9y ago
  Would it be silly to use 'head' and 'tail' then? I learned car/cdr in school so that's what I use.
- currymj 9y ago
  Left and Right are really commonly used in this situation in ML-descended languages. It’s not the worst thing. I don’t mind car and cdr though. I kind of wish they were in Clojure tbh. It feels nice to pay tribute to a half-century of computing history.
  - kbp 9y ago
    > Left and Right are really commonly used in this situation in ML-descended languages. It’s not the worst thing. Are you talking about Either types? OCaml and Haskell at least both call car and cdr fst and snd respectively, but they're rarely used. As for left and right as names, I don't think they're any worse than car and cdr, I just don't think they're substantially better.

lispm 9y ago

Lisp has both CAR/CDR and FIRST/REST. It's usual programming style to use CAR/CDR when working with CONS trees/graphs and FIRST/REST when working with lists.

AnimalMuppet 9y ago
OK, but isn't CalChris's point still valid? Something other than CAR/CDR would have been a better choice for working with CONS cells, even back in 1959. Do I think it's a big problem? Not really. Are the names less than optimal? Yes.
- kbp 9y ago
  > Something other than CAR/CDR would have been a better choice for working with CONS cells, even back in 1959. > Are the names less than optimal? Yes. I don't think that's self-evident. The only better alternatives I've ever heard suggested are something like left and right, which I think are probably slightly clearer names for what the functions do, but not significantly so (I don't think "a cons is a thing with a left and right side" is that much more intuitive than "a cons is a thing with two parts called a car and a cdr"). Further, I think in 1959 when most everyone using Lisp was hacking on the implementation itself, names that are mnemonic for what the machine they were actually using is actually doing make sense; I don't think they're clearly sub-optimal now, but I think they were even less so in 1959.
  - AnimalMuppet 9y ago
    I assert that naming the interface for the implementation is almost always the wrong thing to do. (And then you ask me what would be a better name, and I don't have a perfect answer. "left" and "right" are the best I know of...)
    - kbp 9y ago
      > I assert that naming the interface for the implementation is almost always the wrong thing to do. I think it's perfectly acceptable if there aren't really any better names for the abstraction you're creating. car and cdr work just fine as made-up arbitrary names for an abstraction that doesn't really have any more natural names for them; the fact that 60 years ago they weren't actually entirely arbitrary doesn't really change that.
- kazinator 9y ago
  These names were very well thought out. They were part of a larger vocabulary in fact; most of that vocabulary died out and only car and cdr remained. There were things like cwr, ctr and others. MacCarthy and gang were well aware (of all people!) that the language is machine-independent and not something that will be forever stuck to the particular IBM machine; they used these names anyway because there was nothing wrong with them. By the way, the C and R are not machine specific; any machine can have a "register" and any register can have "contents". On any machine where we implement cons cells, we can just call the two fields registers, understood as being A and D. These, in turn, can be "ante" and "de", if someone desperately needs mnemonics: https://en.wiktionary.org/wiki/ante https://en.wiktionary.org/wiki/ante https://www.etymonline.com/word/de- https://www.etymonline.com/word/de- "active word-forming element in English and in many words inherited from French and Latin, from Latin de "down, down from, from, off; concerning."
  - DonaldFisk 9y ago
    In addition to car and cdr, there were cpr (Contents of the Prefix part of Register) and ctr (Contents of the Tag part of Register). > By the way, the C and R are not machine specific; any machine can have a "register" Slight nitpick: the first machine on which I had a decent Lisp to use (Portable Standard Lisp) was a Burroughs 6800, a stack machine with no general purpose registers, or indeed any registers directly accessible by the programmer.
    - kazinator 9y ago
      Vi has "registers" and so do TeX and troff. I think once upon a time, it was more common to use the word for things beside CPU registers and I/O ports. If a cons cell is a context with two registers (R) whose contents we can access (C), the only vestiges of that IBM machine are the A and D letters sandwiched in between to distinguish them. Compared to how option letters change meaning between Unix commands, it's nothing. -h help? Nope, h)uman readable sizes. ^ to anchor regex at beginning; $ for end. Because on common keyboards, ^ is on the right, and $ on the left!

kbp 9y ago

> So why stick with the names of 704 registers? Was there a purpose to this? Are there other, much better names for the two parts of a pair? First and second aren't great because they're not clear that it's a pair and not a larger sequence (and the difference is important in Lisp; since lists are built on conses, it would be odd to get the tail of a list by calling 'second'). The only other names I can think of would be something like left and right, but I don't think those would be substantially easier to understand for beginners than car and cdr. When working with lists, many people prefer to use the functions first and rest, which behave identically to car and cdr but are more meaningfully named for list applications. They would be terrible names for the general cons-handling functions, though.

lisper 9y ago
I use FST and RST (for FIRST and REST). I like them because they are short, mnemonic, and can be composed like CAR and CDR, e.g. CADADDR = FRFRRST (though I also believe that if you find yourself doing this you're almost certainly doing something wrong).
Steltek 9y ago
Would it be silly to use 'head' and 'tail' then? I learned car/cdr in school so that's what I use.
currymj 9y ago
Left and Right are really commonly used in this situation in ML-descended languages. It’s not the worst thing. I don’t mind car and cdr though. I kind of wish they were in Clojure tbh. It feels nice to pay tribute to a half-century of computing history.
- kbp 9y ago
  > Left and Right are really commonly used in this situation in ML-descended languages. It’s not the worst thing. Are you talking about Either types? OCaml and Haskell at least both call car and cdr fst and snd respectively, but they're rarely used. As for left and right as names, I don't think they're any worse than car and cdr, I just don't think they're substantially better.