Re: [stack] disallowing recursive definitions

John Cowan <[email protected]> wrote:
> William Tanksley, Jr scripsit:

>  > Seriously, I'm confused. Why did you start your paragraph with "no,"
>  > if you weren't intending to disagree with me? This probably all comes
>  > down to miscommunication again...

>  I can no longer fully sort this out, but your remark ended with a smiley,
>  so I assumed it was ironic, and was answering (seriously) that the ironic
>  statement was wrong, meaning that what I took to be your underlying actual
>  opinion was correct.

I consider this fully sorted out. Excellent. The problem is that I
failed to explain precisely why I said what I did, leaving you having
to wonder whether I was being deep and ironic or just shallow and
obvious. Your reading makes perfect sense.

>  In any case, a tail-call in a concatenative language is a possible stack
>  reshuffle to get the arguments in the correct order for the new function
>  and then a GO TO to that function, which is essentially how it's implemented
>  in tail-recursive applicative languages like Scheme: you shuffle the arguments
>  about in registers, possibly spilling to the stack or restoring from the stack
>  as needed, and then do a machine-language GO TO.
>  Assembler: the ultimate concatenative language....

This isn't a useful way to compare languages -- yes, they can all
generate machine code, and given a hypothetical optimizer it'll be the
same for the same semantics. Well, it'll be useful if you can specify
what that optimizer looks like.

A tail call in a concatenative language is a structured GOTO to the
beginning of the function, nothing more. A tail call in an applicative
function is a parameter reassignment combined with a structured GOTO
to the beginning of the function.

>  After fixing the Y2K bug in an application:     John Cowan

-Wm