They might at some point migrate to the Linux kernel as the basis for Windows, yes. I cannot imagine what the advantages would be in the moment of not doing this, other than the resistance within the organization …
Also see this statement:
The company stated that while support for eBPF came to the Linux kernel first, there is growing interest in bringing the technology to other operating systems.
This is really totally crazy eBPF wasn’t really a thing out there and then it just so happened that it came to Linux first. That’s like saying io_uring “came to linux first”, or CFS scheduling … BPF came to Linux. This was a way to efficiently filter network packets. Then it changed so completely that it suddenly became a universal way to run extremely powerful pieces of code with extreme efficiency for a growing list of purposes. It’s now basically a kind of safe virtual machine as part of a kernel that is pretty much (or complete, I don’t mind …) complete, something BPF was never (and never intended to be). eBPF is just “Linux” like io, operations scheduler, or linux namespaces. It may be based in part on or grown from other concepts, but it’s very Linux-specific.
Anyway, as smartbit points out below, Microsoft has taken over kinvolk for their knowledge of eBPF – so they might start working on this. I’m curious, but I think Microsoft will be better off once you integrate the Linux kernel.
