[net2-wg] meeting reminder for 2/13/2009

[Razvan Musaloiu-E.]

Hi!

On Fri, 13 Feb 2009, Philip Levis wrote:

> I can't make the meeting today; I have a day-long obligation.
>
> One comment on the Logger interface. On one hand, the idea of having a
> simple logging interface is good. On the other, having it be a global
> variable (as well as the system time call) is not. One of the points
> of the component model is that it makes all dependencies explicit.
> 
> Moving away from some of the globals in 1.x has greatly simplified
> finding where code is.

I don't think finding things is much easier. :-( Even right now we have in 
the tree wirings to LedsC that are not used. They are harmless but if one 
of them would be use I don't see why would be that easy to track down 
because I'll have to check if all the call to Leds.X are wired to NoLedsC 
or to find all the LedsC that are really used.

One benefit of components (I'm thinking at Leds right now) is the fact 
that they give control at very fine granularity: each use by a component 
can be turned off (by wiring to NoLedsC). But for something that is truly 
global there is no such need though.

> The conclusion core made was that existing C library functions could 
> remain (e.g., pow, other math calls, printf), but that adding extra 
> TinyOS global C functions is a bad idea. There's a reason why there 
> isn't a "Globals.h" right now.
>
> One issue that arises is that using globals requires that every caller
> of the global invokes the same function. For example, you could not
> have some component's logging to go flash and others to the serial port.

If something is truly global then this should not be an issue.

> More broadly speaking, introducing global functions raises the
> question of what functions should be globals and what should be nesC
> interfaces. For example, how about Leds? Random?

For Leds seems a good idea to have control at fine granularity. I'm not 
sure about Random but the local time fits perfectly the definition of 
something truly global. :-)

To conclude: I though this type of logging would be useful to others. I 
was wrong. :P

All the best!
Razvan ME

> On Feb 13, 2009, at 8:54 AM, Omprakash Gnawali wrote:
>
>> Friday, February 13, 2009, 11:00 AM US Pacific Time
>> US: 1-916-356-2663 or 1-888-875-9370 (non-Intel)
>> UK: +44 1793 402663
>> Denmark: +45 4527 5090
>> France: +33 1 58 87 73 75
>> Bridge: 3, Passcode: 8946378
>>
>> Agenda:
>> - Logging
>> - IPSN Tutorial
>> - IPv6 (blip)
>>
>> - om_p
>>
>> PS. I might be a little late.
>
>