Does a type like `[[i32]]` need to be forbidden?

[khimru]

@alice

1. Violating the trait bound makes it not-a-type or an-invalid-type or whatever you want to call it.
2. Once it is not-a-type/an-invalid-type, you are allowed (but not required!) to do things that result in there being no well-defined layout for it.

But there are also:

3. Sometimes when something is not a type because of extra superfluous bounds it maybe easily turned into a type (think, again, about HashMap<f32, f32>: it's a type but why?) and then it makes sense to talk about whether it's good idea or not to include such type into the language.

That's basically, what I wrote in the very first message of the whole discussion:

@khimru

@steffahn

So @Rgoogle: For all practical purposes, especially from a language learner’s perspective, it should be very clear that “[[i32]]” is not a real type, and doesn’t have any real meaning beyond being “something that will produce a compilation error”.

Is it really the best way of thinking about it? It's very easy to imagine Swift-style “dynamic Rust” where [[i32]] is easy, simple, and entirely mundane.

Real Rust have picked a different style compilation way which made [[i32]] hard to implement in it and thus made [[i32]] illegal, but I still prefer to think about things like that as about “something that is excluded from the compiler to not repeat ALGOL68 mistake of producing no implementations and no users”.

That's the only thing that I have said: yes, [[i32]] is not a type, but it's not a type because of matters of troubles that such a type would add to Rust implementation, not because there are some fundamental reason for it not to exist.