Re: [ha] Comments on recent postings

robin.at.acm.org
Sun, 25 Oct 1998 12:27:57 GMT

Mitch wrote:
> So, could I suggest that we go forward with an open mind to processors?
I completely agree. What we have to decide on is the function to be
performed first. We can then implement it on whatever fits the bill.
All we have to be careful of is that we don't design a protocol that
needs huge processing or special resources on which to run.

Sorry if my comments about the proprietary or otherwise nature of the
Stamp seemed to be starting a processor war---far from it---I was already
espousing a processor-independent design before I wrote that paragraph,
and it was actually supposed to show that I had come around to thinking
of the Stamp as non-proprietary anyway (as I can't imagine wanting to
redesign its innards).

> As long as we make sure that Gerber files are posted along with
> schematics, then anyone can have PCB's made.
Yes, this fulfils our open-standards goal.

> If there are enough of us interested in a certain design, couldn't
> we just each chip in to have some boards made by a pro? Also, there
> is probably some entrepeneurial spirit somewhere in the group that
> would be interested in providing a kit of parts (including PCB and
> programmed EPROM/PAL/etc.) for a modest fee.
My thoughts exactly.

> Let's keep everything in the public domain and managed by the GPL.
There seems to be no dissent here.

> ... as I run around the house this morning resetting clocks to
> Daylight Saving Time (I'm in the US), I sure would like a module
> or interface that would manage consistency in time for the dozens
> of digital clocks that I have in the house.
This raises some interesting questions. I am assuming we have a fairly
general packet-communication system at the bottom layer, so we could run
quite sophisticated syncronization algorithms (like NTP) if we wish.
But I suspect that just declaring a ``time'' data type that can be
broadcast by a suitable clock and received by anything capable of
using the value would fit very well with the model of the system that
is forming in my mind.

What if two or more clocks disagree? In the same way, what if two or
more sensors disagree on the state of something, like the openness of a
door, or the presence of a person? Maybe the protocol needs support
for resolving uncertainties like this.

The main problem I see with distributing time is that even if we have
support in the protocol for time data (as I'm sure we will), there is
no way to get most of your typical clock-equipped household appliances
to use it. Very few of them are going to have any convenient external
interface to set or regulate the time. Maybe the video has an IR command
to do it, but the chances of your microwave or cooker having anything
but a manual control are vanishingly slim. I suppose we will have to
be content with all our stuff being synchronised at first, then if we
get to the point of developing automation for microwave ovens we'll
have to replace the manufacturer's control logic anyway (unless done by
simulating button-presses---ugh).

> I've resorted lately to using caller-id boxes scattered around the
> house for that since they, at least, are synchronized to the local
> telephone switch clock.
If my caller-id is anything to go by, the synchronization is rather poor.
The protocol, based on the BellCore one I presume you use, gives only
1-minute accuracy, and even then it's often off by one. But then I am
used to having my time accurate to less than a second---the computers
all run NTP and the clocks receive the MSF 60kHz radio signal transmitted
from the national frequency standard at NPL in Rugby. Possibly a bit far
away to much use:-) But surely NIST or someone does something similar?

The clocks are either small digital or analogue alarm clocks that
I bought from various sources---none of them very expensive (about
$30-$50)---or conventional clocks whose mechanisms I have replaced with
the radio-controlled modules. These look like conventional quartz clock
modules (55mm square, powered by a single AA battery) and cost 17 pounds
($28) from Maplin (a large hobby-electronics mail-order company).

Robin.

-- 
R.M.O'Leary <robin.at.acm.org> +44 7010 707044  PO Box 20, Swansea SA2 8YB, UK