[time-nuts] Lady Heather's Fat Client

John Miles jmiles at pop.net
Thu Dec 31 22:55:14 UTC 2009


> An even better solution would be for the server to actually parse
> the TSIP message from the Tbolt that has the temperature reading
> and for the server to do the PID algorithm and temperature
> control PWM.  The temp control code is not that long or
> complicated.  This method gets net delays, etc out of the
> equation.  It does add the complexity of how to configure the PID
> parameters on the server.
>

That's my plan -- once you're done with that code I'll move it to a separate
.obj file and parameterize it in a way that'll link to both the cilent and
server.

It would be possible to send side-channel instructions to the server to tell
it how to control the DTR/RTS lines on behalf of a PID loop running on a
client.  However, the other big reason to run the whole PID loop on the
server, besides lag and reliability issues associated with the network link,
is that we would otherwise need a way to decide which of several possible
clients has the authority to run the control loop.  Right now all clients
are equal in the server's eyes, and I'd like it to stay that way for
simplicity's sake.

It would be OK for multiple clients to request different temperatures if
they want to, just as we currently let multiple clients issue whatever TSIP
commands they want, but giving direct control over the DTR/RTS lines to
multiple clients would be asking for trouble.

-- john, KE5FX




More information about the time-nuts mailing list