[NTS: need to update to blogging software that allows ‘below the fold’ type posts :)
And a follow up from a small chat at the Berkman Center with Dr. Eslambolchi and Jonathan Zittrain (attending from Oxford.) The chat was moderated by David Weinberger.
Q: What does E. mean by end-to-end?
A: in the 80s, investment was all in the core, because that is where the brains was, and where you needed to invest if you wanted high reliability. In the 90s, lots of push to invest in the edges, which neglected the core. Thinks now we need to invest in both, and they don’t work separately. Voice at edges without QoS in the core doesn’t work- doesn’t scale cost-wise or operationally. Interjection: but the core internet scaled. Response: security- can’t be handled at the edges. Cisco by themselves has 17K different IOSes deployed. Too much diversity to protect at edges.
Zittrain jumps in- agrees that security is the core problem; thinks basically we’ve been lucky that we haven’t had massive destruction caused by viruses. Wants to know how the network can improve the situation.
Eslambolchi- network has global view of entire internet. Have the network have ‘sensors’ that can check on how well the ‘net is functioning. Z: does this mean, say, agreements among Tier 1 ISPs to share data? E: yes, Tier 1 is the level to do things at. AT&T extracts terabytes of data a day at this level, and does analysis. They can see patterns which edges cannot. Thinks also that the network has to be ‘cloaked’. Thinks that BGP routing is broken, because all routing is knowable to the entire world. So if you do routing that is more sophisticated, the network can’t be mapped/pinged. Z points out that this does not violate end-to-end, necessarily, since packets go in at one end and come out the other as directed. Thinks also that there needs to be HW virus-checking (paraphrase, not sure I understood that correctly?)- hardwired to actually cut you off from the internet if you get a virus, until the network says it is OK again. Z asks- if this is all at the Tier 1 level, does this help Tier >1? E answers ‘probably not, at least not directly.’
Weinberger and Z pounced, sort of, on the implications of this hardware- basically gives a backdoor to shut down all edges by governments. Eslambolchi strongly agrees that regulation is bad, but doesn’t necessarily seem to grok (or at least want to admit to grokking) that such a tool could be used to restrict ‘viruses’ like blogging about freedom, etc.
Objectives of original chipset design was purely for the business customers. Not really working on it at the consumer level at all.
E protests that he is not a privacy or legal expert, which feels bothersome- basically takes the approach that ‘I’m just doing what my customers tell me’. Z wants a whitepaper ;)
Z asks a (hypothetical) Q: what if someone wanted to start a global wifi sharing network? What are the implications? A from E: sharing the network is illegal, because the ISPs say so.
And that’s that… maybe some commentary thinking later today :)
 I remember having a crazy, expensive setup to have one ill speaker speak remotely at GUADEC- it’s cool that by next year’s GUADEC it should be pretty trivial to do that with a cheap web cam like iSight and gnomemeeting.