It does seem to work with USB serial devices, but only because those drivers on windows in turn emulate a serial port.

So we have the USB driver pretending to be a serial port to Windows pretending to be a TTY to the Linux subsystem.

It's a bit silly, but then I imagine TTY is a whole lot easier to emulate (even with all the cruft on that) than a USB chipset.

Tangential, but there was a recent HN article (I think) (that for the life of me I cannot find a link to) that referenced a Microsoft DSL for describing finite state machines that was principally developed to implement MS's USB driver. Closest link I could find (which is not the correct one) is: https://www.microsoft.com/en-us/research/project/model-based...

Perhaps it was this one from 13 days ago, about Prolog and Windows NT network configuration? https://news.ycombinator.com/item?id=14046420

No, that wasn't it.

P? https://news.ycombinator.com/item?id=12673739

(Only took me about twenty minutes of assorted searches to find it too...)