Re: some messaging questions

[Andreas Fuchs <asf@xxxxxxxxxxx>]

Today, Matthew Dillon <dillon@xxxxxxxxxxxxxxxxxxxx> wrote:
> Theoretically the target port is responsible for handling threads
> waiting on the port, or for signaling potential waiters, and so forth.
> That part of the messaging code has not been fleshed out yet, at least
> in regards to userland ports.  What we will likely end up doing is
> have a type field in the port structure that tells the kernel how to
> deal with it.

Sounds good; I think it would be good to let the receiving process as
well as the sending process specify how to handle the message on the
respective process's side. Say, for example the receiving process wants
the message to be queued up, but the sending process wants to send it
asynchronously; you'd need that if you implement signals (and signal
masks) using messages, anyway. If you can queue messages,
posixly-realtime signals come out nearly by themselves, too (-:

Thanks,
-- 
Andreas Fuchs, <asf@xxxxxxx>, asf@xxxxxxxxx, antifuchs