Standard IP-over-GSM, just as with the systems that monitor street lights, power transformers, and all the other stuff out there.

Disclaimer: one of the companies I am involved in, Viola Systems does that for the smart grid space, and has done it for motorways and other infrastructure.

Oh we have lots of interesting things - mouse traps, lighting, power control, power monitoring, motion sensors, environmental sensors, irrigation controllers, etc etc. Some of these make more sense with an imp card (eg the mouse trap), some make more sense with an imp integrated (also an option). One reason we're not shouting too much about things is that we're rushed off our feet with all the vendors we're currently working with and so drumming up yet more business hasn't been a priority, though that'll obviously change

There are many devices on their way which are imp enabled. Early next year, when the commercial service officially starts, there will be everything from pet feeders, access control, HVAC (very consumer) to elder monitoring (service-based) to parking lot access control (very commercial)... and a lot more inbetween.

None of this working (well, communicating... they still run code without wifi) when the internet is down is not an issue for these applications; the vendors accept this because the systems they were thinking of were already 100% internet-tied too - no loss there.

The fees are very reasonable for vendors, though. A vendor can end up paying $1/year for connectivity - low enough to bundle into a BOM.

What they get for the money is much more than they could make themselves - reliable, secure, able to update firmware remotely, connection monitoring, server-side agents, etc etc. Even if you're making 100k devices a year, you're not going to be able to build, operate and maintain a service for a comparable amount...

The whole idea of an IoT network isn't stupid; the thing which IMO marks this one out as strange is the ridiculous data rates, though something may have been lost in translation. Max of 1kbit/sec... for what, the fabric or a million devices at once? It's possible this important scale factor has been overlooked in their documentation. The lowest rate of 10 bits per second is unlikely to be useful even for low power applications simply because you need to keep your device powered for a minute to send a header & packet

There are whitespace startups like Neul, who are making a robust wide area solution to compete with cell companies, targeted squarely at IoT. This makes a lot of sense, and they have significant data rates... can't wait to see their silicon as a rackmount transceiver is a bit big for me!

Wouldn't a cellular imp be nice... mmmm....

I'm very excited about getting a load for light switches as it's going to be a lot cheaper than rewiring my house.

Hey, look what I stumbled across:

http://arstechnica.com/business/2012/11/...-when-its-done/

Reading that article to the end, their problem was depending on an accelerometer. What it really needed was a condenser mic. Much easier to hook up, program, and tune.

Or perhaps a Hall effect sensor, wrapped around the washer's power cord. Cycle ends, current drops. Seems easier than trying to do the detection based on a by-product of the washer's operation (such as noise or movement).

Doesn't really work unless you can split the wiring. The return current cancels the load current.

Solved in one of two ways:
1) Put the imp and sensor into an intermediate plug that the washer then plugs into (rather than directly into the wall socket).
2) Open the panel on the back where the power cord enters the appliance, and add the sensor where the wire is already split. YMMV.

From a productization standpoint, the first would be simplest, but then you have to make products that fit various outlet standards, if you want global reach.

Sure except you are now well beyond just a current transformer/hall effect sensor on the power cord and into much more regulated and more dangerous area for the home hacker...

I see the imp is supposed to be more integrated with the device. There are already ZigBee devices that can do the power pack option albeit without the fancy backend/cloud system.

...power monitoring/control plugs are also on their way for the imp, as we've been working with some asian vendors to seed this market. Looking for when the power drops and stays low for a period is likely the very best way to detect a wash being finished.

Meanwhile, if I were to hack up something like this today, I'd just use a photoreceptor taped over the "active" LED on the washer's control panel. Way simpler and more effective than trying to calibrate a motion sensor.

But it would cover up one of the blinkin' lights, so perhaps not to everyone's taste. (un)Fortunately, our washer has about a dozen or so of them.