BPF: what's good, what's coming, and what's needed

Jon, just a note for clarification with regards to ABI compatibility: The context with regards to ""ambiguous plane" between kernel ABI and the "wild west" of the kernel's internals" is on tracing where things can occasionally break due to kprobes nature or tracepoints getting potentially removed etc etc. BTF will for example help here since structure layout can be figured out for a running kernel by a loader or verifier and therefore member offsets adjusted automatically in future to a certain degree so that a tracing program can run on different kernels without a need for recompilation. For other types like networking programs, it's the same rules as syscall ABI meaning programs will keep running on newer kernels. Perhaps this can be clarified a bit better.