Status of Projects/TTX 2/8/2008
* Defining the status of projects * TTX
* Rodrigo (R), UC Berkeley * Arsalan, UC Berkeley * Phil, Stanford * Razvan (Rz), JHU * Om, USC * Kaisen, UCSD
* Collection material from existing projects * Ask David if he is interested in link layer work * Get code for Zigbee and TyMO checked in
Defining the status of various projects
O: For releasing protocols, what kind of steps do we need to take, and what types of requirements do we need to meet, such as documentations, etc. More specifically, we should be looking at some of these in terms of plans for TTX.
P: When something is released, it goes into main release, unless somebody says not to put it there.
O: So what about testing and bug fixes? Do that before?
P: You can create development branches. But still, the consensus is that stuff in CVS doesn't have to be perfect, but needs to be at least usable.
O: Of course, it will have gone through certain testing. I say it put there, because it becomes easier for other people, even in this working group, to see the code.
P: Well, the one thing that I push back on, is that code churn is bad. People don't feel good about the code if there is a lot of churn
-- Lots of details about the semantics of CVS/SVN and checking out software
O: I think the key thing is making sure user documentation is there. Do we require each protocol to provide tests?
R: There should be at least a single example or a test case that people can run.
P: In terms of how much should we test, there isn't a numerical value. If people are complaining that it doesn't work, it should be more tested, otherwise, it's fine...
O: At least one other working group member should always test a protocol.
P: In core, everybody has to write a test-harness and examples, which would allow other people to test the code.
O: Did you guys agree with the classifications for the other work? The beta, and alpha? The only thing we're releasing right now is Deluge, and focus on Dip.
P: Dip is solid.
O: True, but it needs to be tested more.
O: What about the link layer stuff?
A: I'll talk with David Moss and see if we can figure out a direction that it can take, and how to move forward.
O: Zigbee should be checked in relatively soon, and we are moving forward with contributions. The code for 6LowPAN is also checked in, and further contributions are coming.
P: Next TinyOS release is in June.
O: I'll start collecting material from people, and send out some slides for TTX.
R: I have a question about this new thread regarding CTP and LPL. Any thoughts? Is anybody going to use it?
P: It came about because we were wondering how to use CTP over LPL because of the need for setting the LPL values on a per packet basis. This interface is meant to alleviate that.
R: Om what kind of results do you get on TudorNet with CTP over LPL?
O: We get about 80% delivery with about 40-60% duty cycle. The LPL algorithm is not very optimized either, and we're doing a lot of things other than just CTP.
O: What about the intercept interface? Core doesn't need to get involved, right?