Did Safe C++ ever have a full, correct, fully compliant, reference implementation, or was there only (closed-source) Circle as some kind of reference implementation? Circle, as far as I know, is closed-source.
Technically speaking the clauses on either side of the "or" aren't mutually exclusive. You can have a "full, correct, fully compliant, reference implementation" that is also a closed-source implementation!
Well, unless the implication that Circle isn't "full, correct, [and] fully compliant", in which case I feel I should ask "with respect to what?" and "why do you need those requirements?"
Eh, bit of a mixed bag, I think, depending on the context in which the words are used. "Circle" can refer to the compiler/toolchain or the set of C++ extensions the compiler implements, whereas Safe C++ is either the proposal or the extensions the proposal describe. As a result, you can say that you can compile Safe C++ using Circle, and you can also describe Safe C++ as a subset of the Circle extensions. I wouldn't exactly describe the lines as well-defined, for what it's worth.
> There are presumably differences between them, and I do not know what those differences are, and I do not know if those differences were documented somewhere.
They're sort of documented indirectly, as far as I can tell. Compare the features in the Safe C++ proposal and the features described in the Circle readme [0]. That'll get you an approximation at least, albeit somewhat shaded by the old docs (understandable given the one-man show).
> I cannot find any occurrences of "reference implementation" in the Safe C++ draft.
The exact words "reference implementation" may not show up, but I think this bit qualifies (emphasis added):
> Everything in this proposal took about 18 months to design and implement in Circle.
[0]: https://github.com/seanbaxter/circle/blob/master/new-circle/...
I'm not sure about the second sentence either? Circle (supposedly?) implements everything in the Safe C++ proposal, so in that respect Safe C++ exists. Alternatively, you can say Safe C++ doesn't exist because major compilers don't implement it, but that's kind of the point of the Safe C++ proposal (and many (most?) other C++ language proposals, for that matter) - it's describing new features that don't currently exist but might be worth adding to the standard.
> people might be interested in alternatives that they expect might be available sooner.
This is also a bit funny because this was one of the more contentious points of debate in the Safe C++ vs. profiles discussion, and the impression I got is that between the two Safe C++ was generally considered to be closer to "might be available sooner" than profiles.
> and I think it would be fair to consider alternatives at this point
I would assume those who could consider alternatives already have, and that those (still) interested in safe(r) C++ do so because the alternatives are insufficient for one reason or another.
Like, Python and Javascript both have many "implementations", and those are some of the most popular languages. Python does not have an ISO specification. But Javascript does have an Ecma standard, ECMAScript.
Rust is getting another implementation in the form of gccrs. And there is work on a specification for Rust https://rustfoundation.org/media/ferrous-systems-donates-fer... . Arguably not a standard, but still helpful.