Someone mentioned the brittleness of Chord.

Thank you for making the only insightful comment I've seen in all of the web discussion. You are quite correct that Chord is generally considered to be brittle, and so might not work as advertised in the paper.

Claims of Chord's brittleness are entirely theoretical at this point, but certainly not to be ignored. Real world testing is needed. There are other distributed hashtable implementations which could be dropped in place instead of Chord. Kademlia is quite the rage these days in the p2p-hacker community. Also Pastry and friends. These all need further testing as well to know how they work in high churn environments. I also have some hacks to make Chord more churn-resistant. It's an exciting area of research.

Yeah, I concur. Given that it hasn't happen yet (the internet was only ALMOST brought to it knees) we can reasonably say that it NEVER happen as the technology will undoubtedly NEVER get any better.





wait a minute..... if it was ALMOST BROUGHT TO ITS KNEES on monday, just what do you think IS going to happen NEXT YEAR????? (oh, waitaminute, that will only effect "the students upstairs" sending emails to their friends)

Our company link has been down twice over the last 48 hours for a total of about 5 -- all during business hours.

A couple of staff were mildly annoyed that they couldn't access the odd goverment website to enter or retrieve a small amount of info, but otherwise the staff -- doing their work on our local database -- noticed less than the students upstairs sending emails to their friends (or whatever).

Since it often takes less time for experts at a decent ISP to reconfigure a few firewall/proxy/filter things to block a DDOS than it does for any given staff member to actually notice that "The Internet Is Down" I think we can probably cope with any sort of "cyber attack".

Taking out DNS servers is a difficult way to break the net.
Most popular sites are going to be cached already. Sure
some of those caches are expiring all the time, but that's
just going to be a small number unless you can keep the
servers out of service for a long time.

Other attacks could be more damaging.

I don't think anyone has ever succeeded in getting Chord to work on a large scale. It is a relatively "brittle" system in that it requires a very specific topology which constantly must be updated as nodes enter and leave the system. So this is probably not how a successful massive DDOS effort would be architected.

That doesn't mean it won't happen, but it does suggest that the attackers will need a more ad hoc and fault tolerant architecture than Chord, something less efficient than Chord's brittle perfection but more workable in the real world.

Of course, the other day, the Internet was almost brought to its knees by an attack on the DNS root servers, and I never even noticed. So maybe the idea of attacking the net as a whole is harder than it sounds.