[Tinyos-host-mote-wg] Re: [Tinyos-2.0wg] Resource handles

[David Gay]

On 10/21/05, Cory Sharp <cory at moteiv.com> wrote:
> Why might I implement each one as a unique resource user?  Because I
> don't know how many of those different commands to CC2420Control will
> be called before the resource becomes available, I need to either 1)
> force each user of the CC2420Control interface to re-request if a
> command FAIL'ed, or 2) use just one resource and internally implement
> some form of internal command queueing, or 3) make each command be a
> unique resource user, handing off the queueing to the arbiter.

Hmm. I would've made each user of the CC2420 (at this level) be a
unique user. Then it's up to them to get access before calling
whichever set of CC2420Control commands they care about.

> As much as possible, I don't want require components above to check
> for error codes, because that's tedious, error prone, and possibly
> duplicative.  I don't want to necessarily implement an internal queue
> if I don't have to for the same reasons.

Well using request/granted has no error checking (assuming you don't
carelessly request twice using the same resource id, but I'd argue
that's normally a bug).

> And if I want my system to
> be responsive, I want the commands satisified immediately if they can
> be, otherwise deferred.

That's the one thing missing. But I'd like to see the argument showing
that there's a responsiveness problem with the current suggested TOS2
approach (request/granted in tasks, generally at the whole subsystem
level) before changing it...

I also note that your proposal seems to have lots of resource ids,
which costs RAM and CPU in the arbiters. So tradeoffs may not be
completely obvious.

David

_______________________________________________
Tinyos-2.0wg mailing list
Tinyos-2.0wg at Mail.Millennium.Berkeley.EDU
https://mail.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-2.0wg