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

[David Gay]

Gilman Tolle wrote:
> On Apr 11, 2005 10:46 AM, David Gay <david.e.gay at intel.com> wrote:
> 
>>Gilman Tolle wrote:
>>
>>>(discussion moved to tinyos-host-mote-wg)
>>>
>>>So, let's list some problems.
>>>
>>>The only way for a host to discover whether a mote is attached is to
>>>send a data message to a (hopefully) unused AM type, and listen for
>>>the ACK.
>>
>>Actually you could just send a packet with an unused packet type, and
>>you'll get P_UNKNOWN back if there's a mote there (ok, I haven't checked
>>if FramerM actually does this, but it's supposed to)...
> 
> Just as before, we're guessing about unused types. The guessing is
> made simpler by the fact that only two packet types exist, but we're
> still guessing.

Please. If we were to do this, we would then go and reseve a "never use 
this packet type" packet type. Or, we would be going down a "extend the 
existing protocol with some connection-like features" path, and define 
those (and this might be a bad idea, as it might then be better to just 
reuse PPP - but this isn't immediately obvious).

> We have clearly-defined serial links. Let's make sure our software
> knows their status, on both ends. Let's formalize the link
> establishment process.

While I'm all for the results this would bring, I would us to at least 
consider the costs (see next message) and alternatives, and not post 
messages which imply a "the existing stuff can't be fixed, let's throw 
it all away" mentality.

David