[Tinyos-host-mote-wg] Re: TinyOS 2.0 host-mote working group

[Gilman Tolle]

On Apr 11, 2005 10:58 AM, David Gay <david.e.gay at intel.com> wrote:

> One of the reasons I didn't want to go for full PPP is because it's
> connection-oriented. This may not sound like a problem, but consider how
> we typically use base stations: we're constantly changing motes,
> reprogramming them, resetting them, etc. There's no clear time when the
> connection should be established, or shut down -- unlike when you're
> dialing in to your ISP...
> 
> This is of course solvable (timeouts, attempts to reestablish the
> connection, changes to the PC-side programming model, etc), but it's
> extra work and doesn't really match the current usage model, which is
> more a connection-less message-passing style.

Ok, this connection-free approach sounds like it arose from the
testbench phase, when it was a small group of people in active
development. But, how often can anyone really plug and unplug motes in
a deployed environment? I don't think this is "how we typically use
base stations" any more.

By the way, apologies if I came across as implying that the only
solution was to throw everything away. It just seems more sensible to
explicitly pick a packet type for link control and establishment
packets than to do it implicitly by defining a packet type that always
bounces. Plus, if we pick a packet type, we can use it to convey
information about the link being established. For example, the
protocol could be used to discover the MRU on the mote side, or to
decide on the expected link control beacon rate.

Gil