[Tinyos-2.0wg] TinyOS 2.x WG/07.5.2006 Meeting Notes

Vlado Handziski vlado.handziski at gmail.com
Thu Jul 6 13:22:35 PDT 2006


TinyOS 2.x WG/07.5.2006 Meeting Notes
------------------------------

 Agenda
------------------------------


   - beta2 status, net2 next steps
   - Long-term toolchain maintenance (David G.)
   - TEP review process/schedule (Alliance, David C.?)
   - Serial communication (Vlado, from last week)
   - Resource arbitration status/evaluation (Kevin)
   - SID interfaces and HAL effects (Jan)

Participants
------------------------------


   - Phil Levis
   - David Gay
   - Vlado Handziski (note taker)
   - Kevin Klues
   - Ben Greenstein

Next Meeting Agenda
------------------------------


   - Moving TEPs out of draft

Discussion Notes
------------------------------

Phil:

   - New RPMs were on the agenda...
   - Kristin is going to do it for the beta
   - net2 workshop at the end of the week. Goals: specification of
   protocols, header fields, structures...
   - discussion about tool-chain has been started on the list

David:

   - I sent an email earlier...
   - historically tools done by David and packaging by Kristin
   - we need someone to take over tools and packaging since he and
   Kristin are not available
   - there is some interactions, needs cooperation
   - we can split per platform: linux/cygwin
   - tinyos tools are stable but someone needs to make small fixes
   - Vlado raised the issue of the msp430 tools

Vlado:

   - msp430 tools are ok, python changes should be updated for final
   release

David:

   - nesC has been maintained by him
   - task is making package RPMs
   - we need people available on short notice
   - very responsive especially close to releases
   - what about TU Berlin taking packaging for linux?

Vlado:

   - we have done something similar for the eyesIFX kits under contract
   from Infineon
   - the infrastructure is there, last updates done in February when
   contract ended
   - we don't have any experience with windows/cygwin and we are not able
   to take upon that task
   - it can work only if we split the packaging per platform
   - before we give a final agreement, we need to find a student wiling
   to do it

Phil:

   - what about testing platforms. Should the packager do the testing?

Vlado:

   - platform guys should test it, they have the knowledge

Phil:

   - some pretesting can be done, making sure this compiles for
   platform...
   - Kristin is not a developer so this was sometimes an issue.

David:

   - is stable when you package it up
   - so who can do cygwin?

David:

   - will talk with AR about cygwin
   - there are requests for windows distributions
   - we have to keep this on our mind, but no one has volunteered to do
   the work

Phil:

   - next item is TEPs. We need to make a longer schedule for reviewing
   TEPs to move them out of draft.
   - start with 101, 102, 103 and move forward.
   - what is a good rate?

David:

   - one per week

Vlado:

   - one per week if other items on the agenda, two can be done if not

David:

   - we need to make clear that if someone is interested in a technical
   issue in a TEP, he has to take part in the call

Phil:

   - we need to agree on the technical issues.. the rest can be ...
   - next week we talk 101/102
   - depending if Jan is back

Phil:

   - the alliance TEP has started to specify the review process for TEPs.

   - once the chair decides it is ready, it goes to the Steering
   Committee.
   - they select a member that is not involved in the TEP development to
   take the lead
   - talked to David about that.

Phil:

   - Vlado has raised an issue with the serial stack last week

Vlado:

   - on double buffered platforms, the event for putDone is not the same
   as "all data has been sent"
   - we faced an issue when

Phil:

   - Ben do you understand the issue?
   - on msp430, double buffering, you get event when ready to take the
   next byte, not when it is sent.
   - So you can shutdown the UART before it is sent.

Ben:

   - what are the solutions

Phil:

   - separate events
   - have split control

David:

   - how to implement event when you have to loop?

Phil:

   - No way to get event that signals that the queue is empty...

Vlado:

   - export HAL level isEmpty

Phil:

   - problem with HAL not being platform independent

Vlado:

   - this is not a problem

Phil:

   - What about dummy byte in stop

Vlado:

   - other side has to take the garbage

Ben:

   - what about timer?

David:

   - split control in the HPL busy loop using a task

Vlado:

   - this can happen also in normal operation

Phil:

   - sending is fine, problem is interaction with power

David:

   - knowing the transmission is done is useful

David:

   - we can go with stop/stopDone in the midterm but we might not solve
   this

David:

   - how do you tell the HPL that you want this to be sent completely

David:

   - what about split phase isEmpty

Phil:

   - would need to send a task for platforms that don't have double
   buffering

Vlado:

   - we need to make distinction about the semantics

David:

   - all serial comm, queues, makes this distinction

Phil:

   - maybe we should talk over email

All:

   - OK!

David:

   - will check if ATmega if double buffered
   - it is and has HW event for TX empty, msp430 does not, it has only a
   flag

Phil:

   - what is the status with the Resource?

Kevin:

   - I said I will send a list, but this has not been done
   - we here discussed a lot about the interfaces
   - today we set down and went through the interfaces and came with a
   solution
   - the resource interface looks the same
   -

   Resource Controller changes not to have request or
ImmediateRequest<http://tinyos.stanford.edu:8000/ImmediateRequest>
   - now granted goes without explicit request
   - before, idle was signaled and the RC had to make explicit call
   - now granted goes directly
   - when he gets requested he decides if he releases or not
   -

   same happens for
ImmediateRequest<http://tinyos.stanford.edu:8000/ImmediateRequest>.
   The PM will then release

Phil:

   -

   lets talk about
ImmediateRequest<http://tinyos.stanford.edu:8000/ImmediateRequest>.

   - AR says not needed.

David:

   - Henri is using it.

Kevin:

   - I could not understand his code. He is not going back to request if
   the first call fails
   -

   ImmediateRequest
<http://tinyos.stanford.edu:8000/ImmediateRequest>does not carry a lot
of overhead now

David:

   - sounds like a good solution

Kevin:

   - for evaluation we need to talk to Henri

David:

   - goes to EPFL, can talk with him

Phil:

   - it might be in the serial

All:

   - and the radio

Phil:

   - Kevin, can you write this up?

Kevin:

   - yes
   - no overhead if no PM

Phil:

   - ADC/SID alignment issue
   - problem when we want to use them at lower levels
   - SID TEP requires values to be left shifted
   - this is problematic at lower levels when we don't have the info
   - solution: interface should not specify this
   - sensor boards TEP will specify this
   - HAL2 will use interfaces without this
   - Jan looks to be fine with the decision
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.millennium.berkeley.edu/pipermail/tinyos-2.0wg/attachments/20060706/095ed659/attachment.html


More information about the Tinyos-2.0wg mailing list