5 ms·

> The old guy hasn't kept up with the times, and is stuck in ancient (and outdated) ways of doing things - manual management of memory Garbage collectors have

> The old guy hasn't kept up with the times, and is stuck in ancient (and outdated) ways of doing things - manual management of memory Garbage collectors have been around since the 60s with LISP. I find it amusing to see that the first language I was taught in an academic setting in 1996, Caml Light (ancestor of OCaml), is still very relevant today!

arethuza 11y ago
I had to write a garbage collector to get my SK combinator evaluator to run on a Unix mini (HLH Orion) in '88. Glad to see that functional languages eventually caught on though! I wish my application to Linn to work on the Rekursiv hardware had gone a bit further though...
- rjsw 11y ago
  The HLH Orion ran garbage collected Cambridge Lisp from fairly early in its development.
  - arethuza 11y ago
    I wonder if they used their fancy bit-slice technology to change the CPU instruction set to be better suited for Lisp? [NB I am resisting the temptation to comment on the assertion that Lisp is a functional programming language! ;-)]
    - rjsw 11y ago
      I think there were extra instructions for Lisp, it certainly ran the Gabriel Benchmarks faster than you would expect for the advertised clock speed. I only spent a day talking to John Fitch about the implementation and most of that was looking at his bindings to the GUI from Lisp. Cambridge Lisp was written in BCPL which was one of the languages provided with the machine so you probably wouldn't want to completely replace the instruction set.
TeMPOraL 11y ago
Pretty much everything that is "new and hot" was done in 60s and 70s with Lisp and Smalltalk, only smarter and better. Hell, even a day or two ago we had an article pointing out that React is basically rediscovered WinAPI. The old guard that seems to be behind the current trends may in fact be ahead, because they've seen how things went in the previous iteration. As the saying goes, every field learns from its history - except software development.
- yummyfajitas 11y ago
  This is true only for a very narrow set of "new and hot". Lisp and Smalltalk did not "do" the modern, statically typed, inference and category theory heavy languages that are rapidly growing. Dependent types were not "done" in Lisp/Smalltalk. There are also a lot of "new and hot" things which are not programming language features at all - e.g., the lambda architecture, the rise of bayesian inference, and other such things. Even relational databases - so old that it's easy to forget they were "new and hot" (with good reason).
  - yodsanklai 11y ago
    > statically typed, inference and category theory heavy languages that are rapidly growing. All of this is quite old too. As I said, I learned ML as an undergraduate in 96, 20 years ago. Back then, many "hackers" laughed about such languages. Actually, that's why I think a more foundational CS curriculum is a good thing, as opposed to those that are heavily based on technologies.

arethuza 11y ago

I had to write a garbage collector to get my SK combinator evaluator to run on a Unix mini (HLH Orion) in '88. Glad to see that functional languages eventually caught on though! I wish my application to Linn to work on the Rekursiv hardware had gone a bit further though...

rjsw 11y ago
The HLH Orion ran garbage collected Cambridge Lisp from fairly early in its development.
- arethuza 11y ago
  I wonder if they used their fancy bit-slice technology to change the CPU instruction set to be better suited for Lisp? [NB I am resisting the temptation to comment on the assertion that Lisp is a functional programming language! ;-)]
  - rjsw 11y ago
    I think there were extra instructions for Lisp, it certainly ran the Gabriel Benchmarks faster than you would expect for the advertised clock speed. I only spent a day talking to John Fitch about the implementation and most of that was looking at his bindings to the GUI from Lisp. Cambridge Lisp was written in BCPL which was one of the languages provided with the machine so you probably wouldn't want to completely replace the instruction set.

TeMPOraL 11y ago

Pretty much everything that is "new and hot" was done in 60s and 70s with Lisp and Smalltalk, only smarter and better. Hell, even a day or two ago we had an article pointing out that React is basically rediscovered WinAPI. The old guard that seems to be behind the current trends may in fact be ahead, because they've seen how things went in the previous iteration. As the saying goes, every field learns from its history - except software development.

yummyfajitas 11y ago
This is true only for a very narrow set of "new and hot". Lisp and Smalltalk did not "do" the modern, statically typed, inference and category theory heavy languages that are rapidly growing. Dependent types were not "done" in Lisp/Smalltalk. There are also a lot of "new and hot" things which are not programming language features at all - e.g., the lambda architecture, the rise of bayesian inference, and other such things. Even relational databases - so old that it's easy to forget they were "new and hot" (with good reason).
- yodsanklai 11y ago
  > statically typed, inference and category theory heavy languages that are rapidly growing. All of this is quite old too. As I said, I learned ML as an undergraduate in 96, 20 years ago. Back then, many "hackers" laughed about such languages. Actually, that's why I think a more foundational CS curriculum is a good thing, as opposed to those that are heavily based on technologies.