Does this make sense? (In other words, would such an approach be
welcomed?)
I think we talked about this a few weeks back. The answer is a definite
yes. What we should do, IMO, is have GDB depend on the gdbserver code.
That way, we can think of implementing a gdbserver as the first step
towards implementing a native GDB (or seen differently, if you have
implemented a native GDB, then you should have a gdbserver for free).
I think that Pedro also mentioned that the GNU/Linux nat support was now
better in gdbserver as well.
That being said, I don't see this as an obvious task. But I would
certainly welcome it.