By Jack Ganssle
Networked Appliances
Published 5/04/2001
Thom Calandra interviewed Jerry Fiddler, one of the founders and still leading light of Wind River Systems. One extract: "Fiddler's goal is to pave the way for the development of a deluge of smart devices, contraptions like Internet-enabled ovens and video recorders that can be programmed from long distances."
I can parse this sentence a couple of ways. One makes much sense to me: the Wind River, like all of the rest of us, is helping flood the world with massive numbers of smart systems. This, after all, is the very nature of embedded systems.
Jerry, Jerry, Jerry. Didn't the dot-com meltdown show that the Internet is not the only business segment extant? Absolutely I agree that a lot of our products will, for better or worse, have protocol stacks stuffed into them. But do I really want a refrigerator that checks the net to flag health hazards of the instant meal I've just taken from the freezer section? Will an Internet-connected toaster improve either my toast or my life?
Some, maybe most, embedded systems are best left as isolated communications-free controllers. While I do agree that Mr. Fiddler's vision has a lot of merit, I think the real future of our industry is an ever-decreasing visibility of microprocessing. When our curtains contain ten thousand processors monitoring sunshine and adjusting light levels appropriately, when every button in your phone is smart, when even a light bulb dynamically retunes it's operation for greater efficiency, then the era of ubiquitous embedded systems will begin.
Let's remember that even where networking is desirable, it's pretty much impossible today and for many years to come. Sure, tossing a TCP/IP stack, Ethernet hardware, and the like into our products is doable and even getting easy. But. how are you gonna connect that system to the net? Where's the T1 line, the phone wires, the wireless infrastructure? When an embedded system costs pennies - or less - who wants to run what will probably always be an expensive connection to the real world to the thing. especially in those so many cases where the customer-benefit is dubious?
Some claim wireless is the answer. No way. Embedded systems must be reliable. When I urgently need access to the toaster there's no way I'll tolerate "network errors". The one massively-deployed wireless network - cell phones - has rampant disconnects and other tribulations.
Hey, I can't even get a net connection into my home, recently at least. When I moved one slip down the dock (OK, OK, I live on a boat), Verizon moved the phone lines but screwed up the DSL connection. Worse, they can't call the repair department to have things fixed; they, the phone company, must send a letter. Not an email, but a real physical letter (remember those?). It's been three weeks and they're predicting another 3 to go. due to the slow snail mail. Not to knock Verizon too much; they do call, sometimes two or three times a day to let me know the progress (none). For some reason the DSL line was always perfect but the same wires are totally unreliable for a modem hookup, so I'm sneaking around the docks, laptop in hand, furtively looking for better phone jacks.
Somehow I don't think we're ready for a net-centric embedded universe. Till connections are totally reliable, nearly free, and persuasive, I'll bet the future of this industry still lies largely in isolated computers doing their thing quietly.
Mr. Fiddler did get a very important point right, though. As he was quoted in the article: "When the phonograph came out, (Thomas) Edison thought it would be for communication, and when the phone came out, (Alexander Graham) Bell thought it would be for music. We are at the start of a tech revolution and no one knows what it is going to look like."
Truer words were never spoken. Twenty years ago no one could have dreamed that PCs would be so common and so powerful. The embedded revolution still lies ahead.
One survey showed that 47% of attendees at the ESC were doing Internet-aware embedded systems. What do you think?