In his talk about Designing value classes for modern C++, Marc Mutz mentions the concept of inline data (in the 6th slide, when talking about thin abstractions).
I searched for what that means, but couldn't find anything close to a definition of what inline data is (even in the standard, in which I did not find any mention of "inline data").
I suspect the concept may be akin to inline functions, i.e. when assigning to/reading from it, the compiler optimizes away the overhead that comes with the fact that it's a member of a class, though I'm not sure.
So what it is exactly, and how does it work?
Nothing novel I'm afraid. He just means class data that's defined in the class definition, the bog-standard way, rather than being hidden off in a pimpl. The talk is on YouTube and he explains this about 9:30 in.