World's Smallest IPv6 Stack By Cisco, Atmel, SICS 287
B Rog writes "Cisco, Atmel, and the Swedish Institute of Computer Science have released uIPv6, the world's smallest IPv6 compliant IPv6 stack, as open source for the Contiki embedded operating system. The intent is to bring IP addresses to the masses by giving devices such as thermometers or lightbulbs an IPv6 stack. With a code size of 11 kilobytes and a dynamic memory usage of less than 2 kilobytes (yes, kilobytes!), it certainly fits the bill of the ultra-low-power microcontrollers typically used in such devices. When every lightbulb has an IP address, the vast address range of IPv6 sounds like a pretty good idea."
Please specify OS licence (Score:5, Insightful)
At least on Slashdot, it would be nice if posters specified the OSI approved license as it tends to be import for different types of software.
The FAQ says it uses the 3-clause BSD license.
I personnaly like stuff like this to be BSD, while applications are GPL
The IP stack isn't the limiting factor (Score:5, Insightful)
Making the IP stack smaller will not allow low power devices to harness the power of the Internet because while it lowers the bar for technically interacting on the Internet we can't do so safely with a device that can't also implement sane security.
If a light fixture can't execute a secure authentication mechanism to determine whether it really should be turned off/on then it really shouldn't be taking those controls (or reporting its status) to IP queries. These requirements are already beyond the resources needed for less optimized IPv6 implementations this brings us back to Amdahl's law doesn't it... Don't optimize blindly.
Which way is this impressive? (Score:5, Insightful)
With a code size of 11 kilobytes and a dynamic memory usage of less than 2 kilobytes (yes, kilobytes!)
I'm left wondering whether the submitter thinks this is impressively small or impressively large. Perhaps I'm getting old, but to me 11 kilobytes seems rather large. I might be impressed by someone squeezing a stack into, say, 301 bytes, but surely you can implement *anything* in 11 kilobytes.
Re:Sweet (Score:5, Insightful)
And in fact, the Wikipedia page for Contiki links to a web server running on a C64! Shall we see if we can Slashdot it?
Whenever I trash MS-DOS 1.0 on Slashdot, I get a contradictions ("arguments" presumes too much actual knowledge) from people who insist that it's the best OS that could have been implemented on the hardware available in 1981. The counterexamples I usually answer are things like CP/M (the leader before commodity PCs took over), QNX (now sold as an embedded OS, but originally meant as a desktop system), and CTOS (utterly dead now, but my favorite at one time) that all had more power and lower hardware requirements. These examples go right by people because they've never heard of these OSs. (Except maybe CP/M, and then they assume that it's the same level as MS-DOS 1.0, because 1.0 was based on QDOS, and QDOS pretended to be a CP/M clone.) I'm very pleased to learn about Contiki, even though I'll probably never work with it, since it's a prime example that you can even do high-powered OSs with GUIs on 80s-era hardware.
Re:Lightbulb on the internet? (Score:3, Insightful)
Re:in the ipv6 future (Score:3, Insightful)
DRM (Score:2, Insightful)
Re:The IP stack isn't the limiting factor (Score:4, Insightful)
That's actually a really huge microcontroller, on the order of $10 a unit. The cheaper ones, under $1, often have 256 bytes or fewer of RAM. Some don't even have RAM, just a set of general purpose registers and some IO addresses. And RAM is relatively power hungry, which acts as a continual anchor on microcontroller designs. Power is far more important than memory size.
Personally, I think 2KB of RAM is ludicrous for a software stack. But the again, my favorite model has just 4KB.
802.15.4 can be implemented in 100 bytes of RAM. Who in hell needs IPv6 on a MCU? And why in hell would manufacturers add $10 to their unit costs (of say, a $.50 light socket) to enable IPv6 and its attendant problems?
Braindead.
IPv6 as serial numbers (Score:2, Insightful)
I think it would probably be implemented using RFID chips; what the OP was suggesting (I think) is that the RFID tokens (or barcodes, or whatever) use the IPV6 namespace to ensure uniqueness.
Rather than having a separate coordinating authority to hand out blocks of "RFID Numbers", companies would just get a large IPV6 block, and then give a block to each product line, and then to each plant or assembly line, and then give one name to each item as it's produced. They would be guaranteed to be unique IDs and they would have meaning (someone could perform a reverse lookup and get the manufacturer, at least). It avoids having to have a separate, essentially duplicate apparatus, just to manage the separate namespace.
MAC addresses are used this way on some networking gear right now. They have the address printed right on the outside of the box, in bar-code form, next to the model number (and sometimes next to an arbitrary serial number, which is redundant).
In general I think URNs ("urn:bigco.com/model/factorycode/year/dayofyear/serialno") make better unique identifiers than IPv6 addresses would, and remove a layer of abstraction that doesn't need to exist, but the idea is still plausible, I suppose.