@jaybus0: I think we're talking past each other. I never suggested many Things on the IoT didn't have have a reason to be powerful enough to run Linux and be visible to the outside world. Many will.
My simple contention is that most won't. Being on the IoT just means they are assigned an IP address and can communicate by TCP-IP with other Things. You don't have to be a 32 bit processor running Linux to do that.
Indeed, most Things on the IoT arguably shouldn't be visible outside of their local network. A team at the University of Michigan has created an open source, Linux based scanner called Zmap, capable of scanning the entire IPv4 address space. See https://zmap.io/ for details and downloads. It's intended to be a security tool, and discover things which are visible and should not be.
Sure, your car's Infotainment center is a candidate for being on the public Internet. The processor that controls your car's ignition timing isn't, and neither is most of the rest of what the car might contain. The same holds true for for most other Things.
DMcCunney... What you describe looking around your desk is a traditional topology. But people will expect to use wearables, automotive systems, etc. as they would a cell phone, anywhere and everywhere. So not just routers, but many Things will be operated outside of the more controlled local network, or at least that is the idea. Of course there will also be many Things that will and should be constrained to a local network as well, networked sensors, household appliances, industrial test and measurement equipment, etc. The point is, there will be a need for both, and I believe that there will be many, many devices assigned public IP addresses, not just routers. And so a Linux capable device makes sense because of the maturity of the network software.
@Jaybus0: Yes, there are TCP/IP stacks for many devices. There are even stripped down IP stacks like NanoIP. Keep in mind that some of these can only work on local networks. But IoT implies Internet. Many of these devices will need IPv6 support, multicast support, encryption, etc.
Yes, some can only work on local networks. So what? I believe most Things in the IoT will be on local networks.
Looking around at the desk I'm sitting at, there is the netbook I'm posting from at the moment, a desktop multi-booting Windows and Linux, a notebook multi-booting Windows and Linux, two decommissioned rack mount servers, one running Windows and other running Solaris, and a PowerMac running OS/X. They are all on my local network, but my public face is my wireless router, connecting to the cable modem from my ISP.
All of my devices have IP addresses supplied by my router and the router uses NAT to see things get to and from the correct local devices. There are half a dozen machines that might be active and communicating on my LAN via TCP-IP, but only one public IP address seen by the rest of the Internet.
There will certainly be IoT Things that will run Linux and require what you mention, but they will be the equivalent of my router: the interface between their LAN and the greater Internet. The vast majority of Things will not be publicly visible, and I don't see a reason they would need to be.
No, it will not be FTP or HTTP. It will be MQTT, XMPP, etc. For ultra low power devices it will possibly be CoAP (Constrained Application Protocol), seeing that ARM has purchased Sensinode for just that purpose. However, they all work more or less like stripped down HTTP, meaning they are all using text-based messaging over TCP.
Yes, there are TCP/IP stacks for many devices. There are even stripped down IP stacks like NanoIP. Keep in mind that some of these can only work on local networks. But IoT implies Internet. Many of these devices will need IPv6 support, multicast support, encryption, etc.
Also, if we are to connect Everything, then many Things will be operating on line power. I don't trust TCP/IP stacks that have not seen much use as an Internet-facing device, certainly not to save less than 2 W. There is a lot of room in IoT for Linux devices.
Another issue is where this new architecture will actually compete. SoCs based on ARM's M0+ Flycatcher core will not run Linux, although they do hit the sub-50-cent price point for the IoT, including security engines and targeted peripherals.
I'm not sure whether running Linux is relevant. The key in Internet of Things is Internet, which means the Thing must have an IP address and be able to communicate with other Things via TCP-IP. That requires being able to implement a TCP-IP protocol stack on the Thing, but doesn't require the Thing to run Linux. TCP-IP has been brought up on all sorts of platforms.
And depending upon the Thing and what it is intended to do, it may not need a full TCP-IP implementation - just enough to communicate on a specific port using a specific protocol. Will a Thing need to support HTTP or FTP, for example? Unlikely.
A lot of discussion like this seems to assume that all of the Things will be full 32 bit processors that could run Linux, but I don't see that being true for a majority of the Things the IoT will encompass. Intel has its own RTOS - VXWorks - acquired with Wind River, and it's hardly the only alternative for OSes on embedded devices.
"As is, Quark has no future beyond given away for free for educational purposes."
And even there, I do not see it becoming much of a success. I mean if I can get cheaper, higher performance and less power hungry Arduino and other MCU boards from a variety of providers, why would I ever bother with Intel's Quark? The software legacy argument does not hold anymore, wake up and smell the coffee. Another gimmick from Intel if you ask me....
Yes it is quite possible that Intel was talking about some next generation variant in the future. In that sense you'd hope this time it will be less troublesome than the process of making x86 suitable for phones (though 5.5 years after launch of Atom, x86 is still almost non-existent in phones...). As is, Quark has no future beyond given away for free for educational purposes.
That's the thing about this launch though - there are plenty of Intel architecture processors available for embedded boards (including multic-core devices), but this is very deliberately aimed at the microcontroller market in wearable devices and IoT - and, like you, I just can't see it. Maybe Intel thinks the 2W TDP or the 1V core voltage will be enough to get into those designs, or, as I suspect, there is a stripped down version with a much lower TDP in the pipeline - that's the only way I can marry up the public statements with the technology we saw last week.
Blog Doing Math in FPGAs Tom Burke 23 comments For a recent project, I explored doing "real" (that is, non-integer) math on a Spartan 3 FPGA. FPGAs, by their nature, do integer math. That is, there's no floating-point ...