Fabrice, I assume most people were on vacation when you sent your e-mail, so let me revive this thread.
First of, we have very strongly identified that the 6TiSCH group only built on/with standard solutions, otherwise, as you can image, sky is the limit and we never get anywhere. So the remainder of this discussion is of "informational" nature only, to which I'm answering with my co-chair hat off. The problem I see with the approach you are proposing are false negative/positives. I assume you would do some form of CCA at different points in the slot. If I were to commercialize a TSCH/6TiSCH solution, I would really like to be sure some transmission is already going on. And I only "trust" my receiver when I receive a valid IEEE802.15.4 frame which passes security checks. Relying only on some energy detection leads to lots of instabilities because of false pos/neg and opens a HUGE security hole. Just my 2c, IEEE802.15.4 people might throw in some cents as well, Thomas On Mon, Sep 7, 2015 at 2:26 PM, Fabrice Theoleyre <[email protected]> wrote: > Dear all, > > I have a small questions about the 802.15.4 TSCH standard. > > It is written: > > > Thus, a device decrements its backoff when a shared slot is finished. > > /************* Standard *********/ > > Let's assume A and B chose the same backoff value 2 (= a collision occurs) > Then, A selects the backoff value 1 and B the backoff value 2 > > <------ shared slots -----> <------ shared > slots ------> > A |______|_____|_XXXX|______|___ … |_XXXX|______|_____|____] > B |______|_____|_XXXX|______|___ … |______|_XXXX|_____|____| > > Thus, the medium access delay is quite large. > > > /************* Non-Standard *********/ > > I know it is *not* standard compliant. But let's imagine we have rather a > slotted-CSMA *inside* a shared slot (and not *among* the shared slots). > Thus, a device would decrement its backoff after aUnitBackoffPeriod at the > beginning of each shared slot. The backoff would then be paused when the > medium is detected busy in the shared slot, and resumed at the beginning of > the next shared slot. > > We would have: > <------ shared slots -----> <------ shared > slots ------> > A |_XXXX|_____|_XXXX|______|___ … | ______ |_______|_____|____] > B |_XXXX|_____|______|_XXXX|___ … |_______| ______ |_____|____| > > (the backoff is triggered at the beginning of each shared slot) > At a first glance, this non-standardized slotted CSMA would reduce the > medium delay and the number of collisions by selecting larger BE. > > In conclusion, I wonder why was rather the first approach adopted in the > standard? Did I misunderstand something important? > > Sorry for my long email... > > Best regards, > Fabrice Théoleyre > > > > > > > _______________________________________________ > 6tisch mailing list > [email protected] > https://www.ietf.org/mailman/listinfo/6tisch > >
_______________________________________________ 6tisch mailing list [email protected] https://www.ietf.org/mailman/listinfo/6tisch
