Containerization and processes are very heavy weight solutions. At the language level, just using a queue class and closures/anonymous functions is enough and much lighter weight.

Here's one (Statecharts Papers We Love). I will follow up with more stuff about PBP, etc. - if not pls remind me. Today, esp. is weird and I expect to be forgetful ... guitarvydas.github.io/2023/11/27/S...

[meta: I can provide links to my Statecharts slides and PBP stuff if you'd like].

This "other" model can be programmed using Harel Statecharts and my PBP stuff. It can be done and IMO, it's 10x simpler.

That model doesn't need magical, complex ideas like cache coherency by default - and - it fits the notion of geographically distributed nodes better than the memory-sharing model does.

The other model is: using over-the-hood synchronization, like explicit network protocols.

The function-based model that we adopted was one of memory-sharing.

I think that it's deeper than that, although I *know* that using multiple, simpler notations is a good thing.

I'm on a Mac using bluesky in my browser (Safari). I had to chop the response up into 4 parts to make it fit the arbitrary character limits. I posted one after the other. On my screen, it shows the 4 parts in reverse order...

[meta: obviously, I don't yet know how to control and manipulate responses. Read the next 4 comments in reverse order.]

It would be much simpler to grow an IC with 100s of MC6800s on it, instead of bloating ICs with caches and memory-sharing.

(1) The basic model of the innards of a CPU is a single-thread, step-wise recipe, but, (2) the basic model of asynchronous devices is that of many CPUs each with their own private memory (after 50 years, we're kinda backing into that by inventing L1 caches.

Even Smalltalk, IMO, doesn't do message-passing - it does synchronous method calling.

Nope, it's about /not/ basing every programming language on the concept of how assembler works - synchronous, sequential. Function-based programming (and FP) has its place, but that place ain't one of dealing with asynchrony.

One must create additional code to implement isolation using functions (e.g., queue classes or operating systems). CPUs _can_ implement paradigms that differ from the popular function-based paradigm, such as Prolog and Forth.

The brevity of this statement belies its depth. Functions cannot, by definition, isolate data _and_ control flow.