At least in the United States, there’s nothing to stop someone from independently implementing the Wolfram Language and libraries and not paying Wolfram anything. See Oracle v. Google (U.S. 2021), which held that copyright law does not protect libraries (e.g. structure, function names, semantics) and APIs because they’re necessary for compatibility/interoperability. The highest court said that was fair use for Google to completely copy (reimplement) Java and all its standard libraries. Someone else could do the same thing with Wolfram Language and it’s libraries.
Unfortunately I don't think this will happen: The Wolfram Language has a _huge_ library of functions from various domains, which is where most of its strength comes from. If you had the time and money to reimplement all of them, you could just make a competing product with better API. I don't think there is enough commercial interest in running existing Wolfram Language code without Mathematica.
Exactly. Implementing the language is one thing. Mathics does it well. But implementing the vast library is much harder. (Though the essentials are possible as proved by Mathics.) Let alone assembling the countless example data it comes with.
Also nitpick but you can run Wolfram code without Mathematica directly using the Wolfram Engine.
mathics seems really cool. I was thinking of biting the bullet and buying a copy of mathematica since I'm teaching myself higher level math and want a a language that does symbolic computation. Needing to pull a docker file to run seems like a bit of a pain though...
You could also try Wolfram Cloud. Only significant issues a free account has is the inability to upload (so cannot use found notebooks unless you retype everything yourself) and files expiring (becoming unavailable unless you pay) after 60 days. But give Mathics a chance first.
I think that would be rather pointless. From using Mathematica/The Wolfram Language, the cool part was not the language design but the powerful functions. It’s not like wolfram language code is particularly widespread so better/open versions don’t necessarily need the same language. They need the powerful backend.
> which held that copyright law does not protect libraries
Correct me if I’m wrong but it held that this specific case was fair use, not that structure, function name, etc are not copyrightable. While it does give a huge precedent case, code in itself is still under copyright by default.
As others mention, the language itself isn't that important. Like so many other languages e.g. Python, it is basically a Lisp underneath a layer of infix syntax. Most of the work went into what in Python terms would be called the standard library, which includes a stupendous amount of scientific and mathematical code, a lot of which is not so easy to implement.
Prior copyright cases allowed protection for “structure, sequence, and organization” (SSO) of software. Oracle’s basic argument was that Google copied the SSO of Java for Android/Dalvik. The Oracle decision undid that when it came to reimplementing APIs and libraries so that code could be compatible between systems. Basically, now no one can own an API or prevent anyone else from reimplementing and selling a compatible API, at least not in the U.S.
(I am a lawyer, but not your lawyer. Not legal advice, etc.)
