[coyotos-dev] Sleep for interval

[Charles Landau]

At 2:59 PM -0400 9/30/07, Jonathan S. Shapiro wrote:
>On Sun, 2007-09-30 at 11:34 -0700, Charles Landau wrote:
>>  >Be that as it may, the issue at hand is the sleep interface. What would
>>  >you propose as a better interface here?
>>
>>  Easy: sleep the returnee instead of the invoker.
>>  Perhaps what this means in Coyotos is "send a message to the reply
>>  endpoint after the specified time".
>
>Unfortunately, this seems to require that the kernel allocate infinite
>storage.
>
>Some questions:
>
>   1. What should the kernel do if, at the time the sleep() call occurs,
>      the returnee is not in the receiving state?
>
>   2. What about if the returnee exits the receiving state before the
>      wakeup message arrives?
>
>   3. What about if multiple sleeps() are performed by different parties
>      naming the same returnee?

I really cannot answer for Coyotos, but in CapROS these all have 
straightforward answers:

1. If the returnee key is not a resume key, it is ignored.

2. If the returnee exits the receiving state, the resume key held by 
the sleep object (in this case the kernel) becomes void, and no 
wakeup occurs.

3. The first wakeup invokes its resume key, which voids the resume 
keys held by all the other sleeps.

I should disclose that at the moment, CapROS sleep suffers from the 
misfeature I'm arguing against here. I intend to fix it, and at this 
point I'm not aware of any obstacles to doing so, but maybe I just 
haven't thought about it enough.