2 ms·

The idea of conforming a language specification directly to the constraints of a platform is highly idealistic, of course, but nevertheless one which has a lot

The idea of conforming a language specification directly to the constraints of a platform is highly idealistic, of course, but nevertheless one which has a lot of contemporaneous applicability. Yet, the vogue appears to be more inclined towards abstraction. Here, we see the abstract in its most naked form: the owner of the hardware is the same as the owner of the software. They are the first user. So this is quite a poignant item for the HN crowd. Do we need more, or less, of this sort of thing? Methinks, more.. but it may be knee-jerk borne from a few decades of jade.

ajross 12y ago
I'd say not idealistic at all, it's an entirely practical thing. Early LISPs were coded directly to the instruction set and register widths of the 36 bit IBMs they were written on. BCPL and B were likewise untyped languages with the "word and pointer are the same thing" assumption baked in. Fortran and the IEEE floating point standard evolved in lockstep, etc...
Aloha 12y ago
platforms are transitory - portability is still the oracle worth worshiping at. Look at the folks who ported Version 7 Unix to x86 - 35-40 year old code, running on a new platform.
mhd 12y ago
I've only got a very small value of "few" to jump into this argument, but let me provide a (slightly hyperbolic) counterpoint: electrical engineers often are crap programmers. Don't get me wrong, some of the best programmers I've worked with graduated with an EE degree, but they were working as programmers. If you're not, than matters concerning programming require a context switch that often isn't fully made, resulting in a slightly messed up mindset that usually doesn't produce the most optimal results. This is similar to programmers doing UI/UX, which most of the time doesn't even result in a good UI for programmers, let alone mortal users. Not saying that this was the case here, though. When you've got some serious hardware constraints, it's easy enough to lowball your ambitions, even though theoretically something more generic could be made. Not every team has a Niklaus Wirth in it.

ajross 12y ago

I'd say not idealistic at all, it's an entirely practical thing. Early LISPs were coded directly to the instruction set and register widths of the 36 bit IBMs they were written on. BCPL and B were likewise untyped languages with the "word and pointer are the same thing" assumption baked in. Fortran and the IEEE floating point standard evolved in lockstep, etc...

Aloha 12y ago

platforms are transitory - portability is still the oracle worth worshiping at. Look at the folks who ported Version 7 Unix to x86 - 35-40 year old code, running on a new platform.

mhd 12y ago

I've only got a very small value of "few" to jump into this argument, but let me provide a (slightly hyperbolic) counterpoint: electrical engineers often are crap programmers. Don't get me wrong, some of the best programmers I've worked with graduated with an EE degree, but they were working as programmers. If you're not, than matters concerning programming require a context switch that often isn't fully made, resulting in a slightly messed up mindset that usually doesn't produce the most optimal results. This is similar to programmers doing UI/UX, which most of the time doesn't even result in a good UI for programmers, let alone mortal users. Not saying that this was the case here, though. When you've got some serious hardware constraints, it's easy enough to lowball your ambitions, even though theoretically something more generic could be made. Not every team has a Niklaus Wirth in it.