Years from now, this conversation will be looked back on as "The early seed ideas of Blow"s *JAI-OS*..."

Warning: This video has nothing to do with Lumpy Space Princess from Adventure Time

If you titled the video "Jonathan Blow on LSP", you'd get 10x the views from angry web devs.

I've done a lot of programming in Garbage Collected languages, Java based ones and Python being the big ones. 100% the GC becomes a problem for performance, especially when you're doing too much object-oriented programming. But it is very convenient when writing the code. His point on object pooling etc is very valid. Basically, you end up having to work around the GC to avoid pauses, which is a sign that the language is probably not best suited to the problem you're trying to solve.

He has a point that no one solves problems at a lower level.

I was thinking about LISP but thought you forgot the i

At the end of the video: isn't that just WASM? 😅
Although it's quite slow moving, that was kind of the idea especially with WASI, at least AFAIK

13:03 because no hype, no money

It’s funny how much his VM comment just sounds like the JVM or CLR (.NET runtime)… though I understand that his vision is for a bit of a lower-level one with all those extra features (i.e. GC) being opt-in and with slightly different features (the security sandboxing thing). It’s definitely an interesting idea

You’re old-school Jonathan. Love that 👍🏻 From another old-school guy.

FTWDK recent & modern Linux kernels do indeed allow some system programs to run in a sandbox in kernel space , namely the eBPF subsystem which uses an in-kernel-JiT-compiler to run arbitrary (sandboxed) code to route, change & monitor network trafic, e.g. firewalls.

sad to hear he hates Lumpy Space Princess from adventure time

The weird thing about erlang, no memory sharing between threads.  Copy each time between threads.

I used monogame for a while, and the GC thing in C# is real. Even for a simple game, you feel those hits not very far at all into development, and what you have to do to prevent them is arguably harder and definitely less intuitive than manual memory management with arenas or what have you

What would you like to see from a "modern" operating system?

WNT == VMS++ ?!?! 🤯

I had no idea.  That's crazy.

I don't have a problem with garbage collection, I have a problem with using a garbage collector as a mandatory part of the language.
A 'garbage collecting arena/allocator' where YOU control when/how the collection happens and what gets allocated/managed with it makes more sense

i liked the take overall but he notes the distinction in his answer, the differentiating factor is trust, yes, if you can trust the people on the other side of that boundary then doing a much lower level integration yields allot of performance and functionality gains. however, in a world of decoupled teams, the serialization cost is worth it to have a really strong barrier to handle things like errors in the server process so atleast YOUR program can decide how to handle an error rather than being taken down by someone doing a bad `panic!()`

I know "language server protocol" more than I know "liskov sub. principle". 🙃 I'm not sure if my expectations were double subverted.

WebAssembly