4 ms·

I'm a big fan of Mozilla and the Go team, but do recognize this is something Dlang's Walter Bright criticized from the beginning. As I assume all parties know

by axaxs 13y ago

I'm a big fan of Mozilla and the Go team, but do recognize this is something Dlang's Walter Bright criticized from the beginning. As I assume all parties know much more than I do, I wonder what reason both languages chose this route...

Ygg2 13y ago
I'd rather someone tries segmented stacks and determines it isn't a good fit than to just discard it on a comment from something someone said. This way future researchers have more data points, whereas just dismissing out of hand leaves a rather big question mark.
- axaxs 13y ago
  Well, I agree I'd rather someone try it for another data point. However, this isn't just a random internet comment, it's a man who has probably written more compiler code than about anyone of recent times .
  - Ygg2 13y ago
    That probably is true, but lots of great men were wrong about stuff that were in their domain - see Einstein and Cosmological constant.
kibwen 13y ago
Graydon Hoare (Rust's original designer) was well aware of Walter's opinion regarding segmented stacks. In fact, Graydon heartily agreed with that opinion when it came to 64-bit architectures. On 32-bit, however, segmented stacks do theoretically allow many many many more concurrent tasks than you'd achieve merely with multithreading. Therefore, Graydon's opinion was that segmented stacks would remain relevant as long as 32-bit architectures were. Now, for context, remember that Graydon began working on this language (in his own free time) in 2006, when the architecture landscape was much different. And then later in 2009/2010, after Mozilla picked up the language in order to write a web browser in it, one of the primary goals was to be able to target phones, which lag fairly far behind desktops in 64-bit adoption. But now that Rust has a stable release target of mid-2014, tailoring core language features to the diminishing 32-bit demographic is just no longer tenable.

by axaxs 13y ago

Ygg2 13y ago

I'd rather someone tries segmented stacks and determines it isn't a good fit than to just discard it on a comment from something someone said. This way future researchers have more data points, whereas just dismissing out of hand leaves a rather big question mark.

axaxs 13y ago
Well, I agree I'd rather someone try it for another data point. However, this isn't just a random internet comment, it's a man who has probably written more compiler code than about anyone of recent times .
- Ygg2 13y ago
  That probably is true, but lots of great men were wrong about stuff that were in their domain - see Einstein and Cosmological constant.

kibwen 13y ago

Graydon Hoare (Rust's original designer) was well aware of Walter's opinion regarding segmented stacks. In fact, Graydon heartily agreed with that opinion when it came to 64-bit architectures. On 32-bit, however, segmented stacks do theoretically allow many many many more concurrent tasks than you'd achieve merely with multithreading. Therefore, Graydon's opinion was that segmented stacks would remain relevant as long as 32-bit architectures were. Now, for context, remember that Graydon began working on this language (in his own free time) in 2006, when the architecture landscape was much different. And then later in 2009/2010, after Mozilla picked up the language in order to write a web browser in it, one of the primary goals was to be able to target phones, which lag fairly far behind desktops in 64-bit adoption. But now that Rust has a stable release target of mid-2014, tailoring core language features to the diminishing 32-bit demographic is just no longer tenable.