David Culler Part 2

David started with the introduction of the early visions for the internet of the future, with tons of real devices integrated to it. It was commonly argued that the classical OSI layers were to disappear in the area of WSN, because very optimized applications, fully customized for particular purpose were necessary. David argues that this view was wrong, because with recent IPv6 full stacks on devices (I’m thinking of the ones in TinyOs and Contiki) one could extend directly the Internet on devices, without requiring protocol translating bridges. Which rocks, coz gateways will become just routers.

With this implemented, WSN apps get much easier to develop, as you can just ping and telnet your devices, and the actual communication protocol is hidden. As David says with humor “now grad students do in one day what my phd students needed a year to do”. Besides, IETF folks are happy now, as we have finally found an application for IPv6. Oh TinyOs just released their IPv6 stack.

This thing is that all IETF protocols for the Web were done for phat computers, and certainly not low-power devices with *shitty* communication (very variable with lossy links), with modest bandwith, and little frames. On the other hand, the trick is to mix WSN people who know well how to design protocols for WSN with the IETF who know how routing and how to engineer the net. He calls that a cross-fertilization of WSN & IETF ideas, so that WSN common algorithms can be formulated in the routing world, and on top of it you just get TCP and UDP. Also, when one could use IPv6 on such small devices tons of net features come for free (huge adressing space, autoconf & management, bottstrap-dhcp, etc). Full IP support will finally hide the lossiness of the link (ok, doesn’t exist but cool word though). Now devices can really become plug & play, and no more need to put some fixed IP in the firmware when coding devices. He also raised the problem that there’s currently no standard low power MAC protocol, and we need to make one stick (either adapting what’s there or pushing a new solution).

David also asked very provocative questions, I liked “will something replace TCP/UDP soon?” and “can we find a new mac protocol optimized both for WSN and normal computers?”. Excellent and relevant questions. Even if my notes are totally disorganized, I found David’s talk extremely well structured and founded. As Dom said, we liked it a lot, and he was totally cool encouraging us on our pursuit! The more time passes the more motivated WOT (yes this site).

I’m currently totally looking into IPv6 for these devices (tmotes) and comparing “real”REST over IPv6 and optimized mini-REST in Tinyos and contiki, and can’t wait to see how all this will work. Oh when will they release a fully IP stack on sunspot??? We already got a REST-server, and my freshly arrived router is totally looking forward to tell nasty things to my spots… ah these cute electronic beasts. Soon, you’ll be able to surf with your iphone my spots and turn of my lamps with it. Would be cool… A couple of weeks…. Ok I go now… Oh Just to make you even more jealous: I just received my chumby. YEAAAH. Too many toys, and too few fingers to code them all…

You may also like...