[Tinyos-devel] Message_t thoughts: prototypes

[Miklos Maroti]

Hi Eric,

On Sun, May 24, 2009 at 3:50 PM, Eric Decker <cire831 at gmail.com> wrote:
> On Fri, May 22, 2009 at 2:42 PM, Miklos Maroti <mmaroti at math.u-szeged.hu>
> wrote:
>>
>> Dear Eric,
>>
>> You brought up very good points, and I am very happy that others are
>> also engaged in the discussion. At this point I am convinced that both
>> Jan's and my proposals could be made to work. But to really compare
>> the two approaches we need use cases, very well defined used cases.
>
> I agree.  To really flesh out the ramifications and to really understand how
> the flow comes together there is nothing like an implementation to help.
>
> I've got a simple yet fairly complicated application that switches
> dynamically between radio and serial (depending on docking state) that is
> fed by numerous sensors producing several streams of data that is also
> intermixed with a couple of other streams of data.  This has multiple layers
> of queuing.
>
> I'm currently morphing this implementation around to clean some things up
> because I've run into the how do we encapsulate (issues with the way AM is
> implemented) when we don't know how we are going out of  the box yet.
>
> Anyway, I've got an implementation platform that I'm currently using to
> prototype some of the ideas we are working with.

Since the message format for the serial is the same as for the radio,
this switch can be implemented by a single metadata field. No problem
here as far as I can tell. In fact, I might do a serial stack for the
IRIS platform using the exact same layers as I use for the radio, so
at that point we could change the physical layer with a single switch.

Miklos