[Tinyos-host-mote-wg] [Tinyos-2.0wg] 09/21/2005 Telecon Notes

[Kristin Wright]

-------------- next part --------------

Notes from 09/21/2005 TinyOS 2.0 Working Group 

Phil Buonnadonna (Arched Rock), David Culler (Arched Rock),
Ben Greenstein (UCLA), Phil Levis (Stanford), Lama Nachman (Intel),
Kristin Wright (UCB)

09/21/2005 Agenda
------------------
- status
- next pre2 release milestones, schedule
- TEP finalization process
- TEP 113: Serial Communication 

09/28/2005 Tentative Agenda
------------------------------  
- microcontroller pwr mgt
- tep organizations

09/21/2005 Discussion Notes
----------------------------------------------------------

Status
--------
- CC2420 (Phil):
  - by todays' end: acks working; need to do some bug fixes
  - recommend that next release has a non-split-phase CC2420 stack 
- msp430: 
  - phil: is someone doing the msp430/cc2420 stuff? 
  - no one in the meeting committed to this

Going forward:
---------------
- mica2 & micaz: potential milestones for pre2 release2
- culler: key to getting more platforms is to deal only with some of
  the subsets; might be prudent to return to original charter of 
  dealing with hardware diversity; quickest way to get other platforms: 
  get working some subset of the existing teps
- phil: getting cc2420 on telos -- only a handful of components but 
  someone has to write the code
- dgay: the missing platforms are not missing things that we can 
remove from the subset; many undone components essential
- culler: did phil get any insight on other pieces while doing cc2420 stack?  
  - phil: did incorporate resource arbitration into stack; no problems;
    there are many calls; if you look at it as a state machine and then
    add number of events that can come in, you get an exploding number of 
    possible states. Found that the timer system setup nicely; one tricky
    part of timer system: some implementations underneath don't let you know 
    what the granularity is; interfaces representing interrupts
    works well and allow you to flexibly plug things in; overall pretty easy.

tep113 - serial communication
----------------------------------
- Phil sent a current copy of tep113 for convenience:
  https://mail.millennium.berkeley.edu/pipermail/tinyos-2.0wg/2005-September/001145.html
- General comments: 
  - dgay: 
    - overview in section 2; not clear which part of 4-element 
      stack is doing what
    - needs mention of baud rate
    - 3.3: perhaps too much implementation in there; not sure it belongs
    - why SerialP if not public? 
    - 3.4: explanation of data-link length and upper-x length not clear
    - 3.5: too much implementation (serial active message P)
    - 3.5: address ad group not checked so not clear what value is going to be in there
    - would like a brief explanation of cc1000 packet/802.15.4 packet
    - 3.6: crc starts at seqno byte but needs to start at protocol byte
    - seqno isn't a seqno if you don't have a cnxn established; in old protocol: a
      unique identifier
    - old version: return invalid packet (invalid number) ; reserving an always invalid
      type (or ping) 
    - some want to do dma on uart; outgoing you can sort of do that; should there
      be some low-level 
<i had to step away from the phone here and missed part of the conversation. -kw>
- dgay: phil and ben: can you figure through how you can do windowing? api could be made
  transparent on mote side. 
- phil: intended that you'd be ableto do it, but hadn't thought it out fully yet
- dgay: would feel better if you had something in rough form
- dgay: david: thoughts on where pc-side interfaces should be documented?
- david: we're getting several things that are important, useful, but outside of
  tos2 charter
- dgay: serial needs to be in tos2 charter since we distribute the server-side tools
  with tos
- culler: teps describe interfaces, some cases components, collecting all things related
  to a 2.0 release 
- phil: tep1 says all these are documentary; docum: component that is in tos and works; bcp
  tep (hardware abstraction levels); 
- culler: perhaps infomrational and maybe they graduate; can imagine 2.0 impelemn that didn't
  have the tools. 
- dgay: current serial tep should be enough to allow you to build the pc-side tools
- ben: still comes down to the java (that is, the description is language-dependent to some
  extent); not the right way to specify; perhaps use xdr / some protocol level 
- ben: the place where you draw the line is wire-format, not described anyplace 
- dgay: is described, but not sure the place is right
- dgay: connection establisment would affect all such tools regardless of what the server-side pc
  language is so worth talking about
- phil: ben and lama will look at dma side to see if receive-side structure needs to change or
  if it's infeasible without copies
- dgay: thinking a flush would be reasonable
- dgay: automatic baud-rate discovery

milestones
-----------
phil: 2420 and micaz; would be nice if telos family but can't commit to it
  - can chat with cory
phil: can arched rock do telos? 
  - culler: 
    - willing to put work into it sorting out where they can push
    - part of trying to narrow down ; will have a clearer view
      soon on which parts they can contribute
phil: imote2: lama?
  - lama: think that would be arched rock (medium term)
phil:
  - when cc2420 stack ready, maybe we can do a mica/micaz release
  - would really like a nesdoc; separating things in directories wasn't
    good -- too many directories; need nesdoc
    - dgay: woudl someone like to work with xml/nesdoc?
    - phil: know nothing about nesdoc, don't think we can commit to that right now
    - phil: might be able to find someone and will synchronize w/dgay on that
    - dgay: if no one found, will do it; won't commit to a date until he gets 
      more information
- tu-berlin: vlado on vacation; jan couldn't make call 
- other milestones?
  - storage, power management, higher-level apps
- phil: good, but let's not get too many; believe pwr mgt high priority;
- dgay: after we get cc2420, can address; routing also high-priority then drips,
  deluge, etc.

process of moving teps out of draft
------------------------------------
dgay: 
- teleconference vote not good because attendance not reliable
- email out to wg, if no objection after a certain time period then
- not clear on what form teps are supposed to take (top-down vs bottom-up;
level of detail); should nail that down before they go out

next week
----------
microcontroller pwr mgt
tep organizations







-------------- next part --------------
_______________________________________________
Tinyos-2.0wg mailing list
Tinyos-2.0wg at Mail.Millennium.Berkeley.EDU
https://mail.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-2.0wg