[Tinyos Core WG] TOSThreads TEP

Philip Levis pal at cs.stanford.edu
Tue May 20 16:19:46 PDT 2008


On May 20, 2008, at 4:15 PM, Miklos Maroti wrote:

>> Of course, function pointers wouldn't be necessary if we
>> supported only Static threads (using terminology borrowed from the
>> TEP), to allow flexibility in the definition and use of dynamic
>> threads, function pointers are necessary.
>
> To be frank, I would not allow dynamic threads at all because of the
> following simple reasons:
>
> - Why would threads be any different from memory allocation, we have
> static memory allocation, static composition, etc.
>
> - You cannot do whole application analysis (included WC stack usage)
> if you have dynamic threads.
>
> - Janos Sallai has a tool (documented in his thesis) that can compile
> away static threads into normal tinyos event driven code. Keeping the
> threads static allows other people to implement threading in a
> completely different way: competition is good (TM).

I think there's an important distinction to make between "help people  
write robust code" and "force people to write robust code." Honestly,  
I think that history so far has shown that while once you really  
understand TinyOS its mechanisms let you write very robust systems,  
initially it is very difficult and there are lots of problems. Let's  
give people a path to robust, event-driven systems that doesn't  
requires a 100 meter plunge.

Phil


More information about the Tinyos-2.0wg mailing list