I remember reading that google tried to impress on WG21 and 14 the need to update the ABI and they were utterly opposed. The result is google is no longer interested in C++. Apple I think is also no longer interested in C++. I think also Swift now has a stable ABI.

A good write up of that is at https://cor3ntin.github.io/posts/abi/

It's implied if you know, but Corentin's post doesn't actually link Titus's C++ proposal P1863 "ABI: Now or Never" which is what we're most concretely talking about here.

Titus correctly predicts the committee's actual response and highlights its danger, summarised in the title. You can pick now, as Corentin desires. You can pick never, which Corentin despairs at - not unreasonably because it means you're giving up performance. But Titus highlights the third option, the committee can instead dither forever never having the courage to announce an ABI break but also lacking courage to declare absolute stability with the encouragement that brings for legacy systems.

ABI Now is - at least in performance - competing with Rust. An ABI swap can make some C++ have the same performance as the analogous Rust which wasn't possible with the old ABI and that matters for outfits like Google.

ABI Never is a different niche. Guaranteed ABI stability gives C++ certainty. It makes C++ a stronger contender for some applications where today it can't go and nor can Rust because people don't think "Just recompile" is a reasonable choice, whether they are correct or not.

ABI Dither is neither of these things. There is no certainty, just because the committee is dithering today doesn't mean they won't make a decision tomorrow, or next year. But meanwhile you're not competitive with the best in class alternatives

Well it's not really C that fills that role it's the C ABI which any language can use without an ounce of C. Rust can use the C ABI.

It would be nice if there was a somewhat higher level ABI that languages could use though. The C ABI is very low level and tedious.

Carbon is performance-first, so ABI instability is the default. We plan to have opt-in ABI stability, but that story is not well developed yet. Replacing C as the lingua franca is not a goal, at least not yet.