Flutterby™! : Skepticism on the Internet of Things

Stemming largely from a talk I gave at an early Personal Clouds gathering, my Twitter feed is full of optimism about #IoT, the "Internet of Things" (and "M2M", which I've remarked on before).

And for a while I shared that optimism. There are so many places in my house where little improvements could make life more pleasant. I've probably remarked before on how a slightly smarter bathroom light switch could allow some heuristics to make me do a little extra work before turning on the bright setting in the middle of the night. I'm still not totally happy with our off-the-shelf fan+thermostat controller solution. On the "I'd like to get to it some day" list is RFID or smart phone activated front door locks so that when we come home with arms full of groceries it's easy to get in.

However...

A couple of years ago I dug an irrigation system into the yard. The promise is that you can set the computer and walk away from it. The reality is that you have to continually monitor the valves to make sure that they haven't exploded, that the solenoids aren't stuck, that the gaskets haven't worn out or mis-seated, and if they have, you have to take the parts down to the hardware store and hope that they still have fittings that work, and that you won't have to re-plumb because someone decided that the in-feeds weren't working as female threads, and now they have to take male threads.

Overall that project has probably saved time, but the maintenance load it added in return is non-trivial. And Charlene still ends up doing a lot of tuning of the irrigation by hand. Set it up and forget it? No.

We have a Toto Carlyle II toilet. We got it specifically because it was set up to take the Toto Washlet seat (which we've also added). The problem is that the Toto float valve assembly sometimes sticks, and the real problem is that nobody carries Toto replacement parts. The toilet is pretty darned good for the water usage, but...

We were also early adopters of Compact Fluorescent Lights. They're great, except that despite all of the hype (and that they're allegedly supposed to still give off 80% of their rated light output at 40% of their expected lifespan) if you don't run around and change them every two years you end up wondering why the hell the house is so dark.

The Nest thermostat sounds great, until, well: RT Matt Haughey ‏@mathowie

My house was w/o heat today because my Nest thermostat's internal battery died. Silicon Valley, always innovating new ways to ruin xmas.

Similarly, over the years friends have spent a lot of money on X-10, or what-have you, and...

The problem is that with every change we make we introduce additional maintenance load. We have additional systems that need to be checked for failure (and anyone who's worked in systems operations for a company with a sizeable quantity of servers can talk to you about dealing with the deluge of alerts that come from monitoring). We have additional systems that need regular maintenance.

And it takes quite a bit of effort to install technology in a house. Crawling through insulation, reworking wiring, all of this is a pain in the tail.

In a world where it's totally reasonable to see service life of decades (paint jobs, roofs, most plumbing and electrical), we're introducing products with lifecycles measured in tens of months. And we're doing so with protocols that don't have any promise of interoperability with previous generations, when the wave of the future comes through again in two years we're back to opening walls and re-wiring.

And for those things that don't, we're out replacing batteries.

Although I was initially optimistic, I now expect that the "Internet of Things" will have the same legacy as all of those in-home intercom systems that were standard in new subdivisions back in the '80s. Over-priced gadgets that get abandoned for the tried and true.

So the big obstacles we need to overcome for the Internet of Things to come about are:

• Reliability. X10 was notorious for not working. Irrigation valves are pretty established, and I still have a few fail every year. I want to know that I'm going to get decades, not months, out of my new toys.
• Monitoring. There are a gazillion people wanting to sell you the "Operating System for the Internet of Things", and it's great that you can point-and-drool your way through scripting operations, but what's the Munin? How do you manage failures?
• Speaking of "how do you manage failures?", what's the fallback? In response to that Matt Haughey ‏tweet someone suggested manually tying the wires together. Actually, this makes sense, and if I re-mount our thermostat (unlike the Nest, ours just uses a couple of AA batteries that get replaced every few years) I'll probably put in a toggle switch so that when the thermostat fails at an inconvenient time we can easily turn on the heater (and, probably, another toggle switch that takes the thermostat out of the loop so that we can more easily turn it off). In the case of the light switch applications I've got, there needs to be some sort of centralized controller making decisions based on what other lights are on or off and what time of day it is. How do we wire up the fallback so that even when the house server crashes I can have light in the bathroom?
• Security models. A year or two ago we thought that partitioned spaces, NATting routers and the like, was enough. Now we're discovering enough viruses running on embedded devices that it's plain that we need better network monitoring to proactively partition networks and watch for traffic that shouldn't be there. And we now know that NSA workers are using the network to stalk ex-spouses; how do you feel about the possibilities that someone else knows how your house is configured?
• Upgrade paths. How do I assure myself that the Saturday crawling through the insulation in the attic won't have to be repeated every year or two?

The problem is no longer cheap switches or sensors, we're going to have to think beyond the simple technological problem if we're going to actually have such a thing implementable.