[Tinyos Core WG] 10.25.2006 Meeting Notes are on the wiki

[Jan Hauer]

In the telephone conference we agreed (preliminary) to allow arbiters
to return SUCCESS to  a client that calls Resource.request() even
though he already owns the resource. On service level (abstracting
from the Resource interface), however, we want a second call to
Read.read() to fail if the first one was successful and readDone() has
not been signalled.

I was about to change the ADC HIL implementation, but there are some
bigger changes necessary than I had expected - I have to keep state (a
bit) for all clients now: when a client calls Read.read I must know
whether he is the owner or not. But setting, say, an "owner" variable
to the client ID when I get "his" Resource.granted() is not enough,
because in between the client might call Read.read a second time. If
he does, I cannot tell that he has called it successfully before (and
not released it) unless I keep state for him [when the ReadNow
interface is involved in conjunction with the Resource interface it
probably gets more complicated]... I may well have overlooked
something obvious, maybe someone sees an easy solution - otherwise I
suggest to change the arbiter semantics rather than introduce the
changes on the service level. Comments ?

Jan


On 10/27/06, Martin Leopold <leopold at diku.dk> wrote:
> Hi All.
> My notes from yesterdays meeting are available on the WIKI:
> http://tinyos.stanford.edu:8000/TinyOS_2.x_WG/10.25.2006
>
> Seeing the agenda for the first time it seems that I missed some of the
> finer points (the motel in Yosemite did not provide the luxury of
> internet =). Feel free to fill in where my notes are wrong, incomplete
> or make no sense.
>
> --
> Regards Martin Leopold.
> Dept. of Computer Science, University of Copenhagen
> http://www.diku.dk/~leopold
>
>
>
> _______________________________________________
> 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
>