Will Crichton

"In a language for learners, we still believe [dynamic scoping] is the correct choice. If a variable exists, it's visible."

A version of Logo was developed for Atari, which apparently was once in the PC industry. They even had Alan Kay as chief scientist! Alan Kay said about his departure from Atari: "I guess the tree of research must from time to time be refreshed with the blood of bean counters.''

Bjarne Stroustrup now presenting the modern history of C++. It's his third paper at HOPL, the only language to appear three times! Testament to the enduring legacy of a time-tested language. (Caveat: I have sworn to never use C++ again. )

"Typical request: * Simplify C++ * Add these two new features * Don't break my code"

Q: Thoughts on dialects, e.g. C++ w/o exceptions? A: I really dislike dialects, because you can't communicate or share. Dialects usually ban features, and that's wrong, the complexity just goes elsewhere. Some of the fastest code I've seen carefully uses exceptions.

Q: Thoughts on Rust and Zig, any influence on C++? A: From what I've seen of Rust, it doesn't fit into my view of C++. I experimented with borrowing ten years ago and found the code got too complicated. We can get type and resource safety without it.

Rich Hickey now discussing the history of Clojure. I've never used Clojure, but I expect to emerge with strong opinions involving words like "complect".

Prediction was incorrect, the right opinion-word was "concretion".

Q: Thoughts on Typed Clojure and using dynamism to manage complexity? A: I don't think C++/Java/etc types are rich enough. If your program is imperative, your problems will be dominated by state. Types cause code to be far more specific and far less reusable than it ought to be.

Q: How did you decide on terms like "concretion" and "complect"? A: I love the dictionary Eg for complect, I was trying to give more specificity to an old notion, coupling, that system designers understand intuitively.

Btw, I'm not covering every talk in this thread (time constraints, and POPL deadline in 2 weeks lmao). But the papers are all freely available here: https://dl.acm.org/toc/pacmpl/2020/4/HOPL … And the video talks + Q&A will be posted eventually! So be sure to check them out.

"APL is a notation as much a programming language, and it's important for subject matter experts to be able to use APL on the whiteboard and argue about it. I've seen open source APLs lead to fragmentation, which reduces the value of a notation."

Stefan Monnier and Michael Sperber presented the history of Emacs Lisp. One focus: modern Elisp has two distinct dialects, dynamically scoped (originally) and statically scoped (added later).

"Back then, dynamic scoping was considered efficient. Lexical scoping seemed kind of expensive to implement. Lexical scoping requires the creation of closures." Same underlying reason as Logo, but different conclusion (performance vs. learnability)https://twitter.com/wcrichton/status/1406999997527191552 …

Q: Advice for systems adding dynamic scoping, eg React contexts? A: Picolisp creator said: "Modularity is the ability to reach into a system's guts and make it do what you want." Unconventional, but not wrong. Dynamic scoping enables users to make Emacs do what you want done.

...The trend at confs like ICFP has been unbreakable abstractions. But systems that result from that, esp editors, are not as flexible at runtime like Emacs. But the balance is hard to find and keep the system comprehensible.

"When I got to Microsoft, I was told in no uncertain terms that the word 'compositional' was not in the MS vocabulary. And we have blown that apart at MS."

Q: Aspects of C#/.NET that are hindering F#? A: The big way we suffered is that F# did not have a strong Linux story during the rise of Linux and subsequently the cloud. That's crucial.