The youtube series is quite good for this, especially the first few episodes where he uses NASM to compile from Python instead of compiling Porth with Porth.
I think it was basically this dude building an LLVM compiler for Porth, which made Porth's creator (Tsoding Daily) uncomfortable because he felt like he was on a race to see who made his own unfinished language better. He says individuals with an "s", so I'm not sure who else he might be talking about, but I remember this LLVM person being the big shocking reason why he initially closed the source code.
As far as I remember, the LLVM dude apologized in a public comment on the same video where Tsoding complained about this, claiming he was closing the LLVM repo and explaining it was just a toy project he made to understand LLVM better that got some attention.
While I agree, I can also see where the creator is coming from here. He is working on making his very unfinished language, and someone else comes and starts making an optimized LLVM version of it. The creator certainly doesn't want the language to always look like it does today—after all, he is still actively working on it, adding new features every week. But he is afraid that the LLVM version will be so much better that people will use that instead of his work-in-progress compiler, and so they will be stuck with an old version of his language.
I think he should have handled it better. Maybe explaining why he didn't want a fork (he may have done this, I haven't watched his porth series specifically), and maybe asking the guy to put out something in his README.
Freezing the public version while continuing to work on it in private but not allowing anyone to see what you're working on... doesn't seem to me less likely to create a fork...
Either people are going to stop paying attention to it entirely (is that a desired outcome?), or they're going to mess with what's public, right?
I don’t agree at all. Unless the author enforces it with license or some legal means and/or the LLVM version author was rude, this is how competition works in a good way. And to be fair, performance is not the only reason people pick between different implementations. A good example being CPython: there are plenty of JIT-enabled python in the wild but I don’t see people turn away from CPython
This is what can happen when you design and implement things in the open. And it’s completely fair for it to happen as long as the design is open source. (Can it even be closed source…?)
But the author responded by temporarily closing it to the public. Which is also OK. He clearly didn’t want to deal with the consequences of openness.
As far as I know, Github generally keeps up with Gitlab. For a while, I was on Gitlab because private repos were free, but since Microsoft purchased Github, Github has that too.
Gitlab's own page about the differences between the two is somewhat biased[1]. Most of the features where "Gitlab wins", are niche and nice-to-haves at best. Mostly "scanning" features, and I'm sure they aren't free. Meanwhile, even Gitlab seems to recognize Gitlab is superior in at least one way (CodeSpaces).
A compiled, number crunching dialect of Forth with a very good vocabulary (and the accompanying semantics) for n-dimensions matrix calculations sounds pretty neat. I’m surprised no one jumped in with a an APL reference, however there are some significant differences between an array processing language and a concatenative language with Fortran semantics.
Dr. Julian Noble (RIP) went down that road but realized it was overkill and switched to providing a "formula translator" for Forth. So it takes infix notation and translates it to RPN Forth code than is compiled interactively like normal Forth.
That reminds me of some code I came across one time which was a BASIC implemented in Forth. It was less than 100 lines of code, which I thought was impressive.
I believe the first version of that interpreter was written by Charles Moore himself.
"Lines" in Forth are rather fluid but I have a version here that is 154 lines when formatted as a text file rather than 1K disk blocks.
