I'm not going to cast an official vote, since I'm not an official
contributor (just lurking here), but I'l de-lurk to point out that I do
not consider the proposal to be intuitive. In general, "object" !=
"thread". From a user's point of view, the constraint that "each stub should only be
used by one thread" is neither natural nor convenient in a general multi-threaded
program. I
have written multi-threaded distributed programs using other technologies
(and I have conributed to other RPC technologies), and used stubs from
more than one thread; the proposed constraint would just be an added pain
in those programs.
