[coyotos-dev] Re-thinking FCRBs

[Marcus Brinkmann]

At Fri, 23 Feb 2007 12:00:44 -0500,
"Jonathan S. Shapiro" <shap at eros-os.com> wrote:
> In Coyotos this doesn't actually require a full thread. It requires an
> endpoint. This endpoint is provided by the selecting process to the pipe
> server. What the pipe server needs is the ability to either (a) store
> this endpoint, or (b) enqueue this endpoint on a receive queue.

That's a neat trick, I didn't think of that one before.  I can't go
through all the details right now, but it seems very plausible, at
least in the specific use case scenario discussed (and as I said,
that's the only one I could come up with).  It works because there is
only a fixed number of possible different types of requests that can
block which have to be differentiated, so the request does not contain
state beyond what can be identified with the EndpointQueue.

Thanks,
Marcus