04 Sep 2004


Just a round-up of SIGCOMM Day 3. I'll just talk about what stood out in my view. I missed most of the first two sessions since they talked about things that I can have no influence with. They were Router Design and Congestion Control. The thrid session of the day, Naming and DNS was quite interesting.

The first paper (Impact of Configuration Errors on DNS Robustness) was interesting. It talked about how they went about detecting DNS errors and then figuring out ways why it is broken. So 2% of the entries were broken, like lame delegation etc. So there was a situation where the DNS servers were down, but the actual host is up. I thought it was quite a good analysis.

The second paper talked about how to apply distributed hash-tables (DHTs) to DNA and how a system like Pastry would work under it. It seemed quite reasonable, except it takes 2 hours to converge in the worst case. Some points that were brought up was the methodology used to compare this new DNS called CoDoNS with legacy DNS. For instance, their DNS ran on internet2 whereas the current DNS is quite slow. There wasn't much about cache invalidation, but I should read the paper more carefully on how that is handled.

The third paper was a position paper about what to do with the next generation naming system. They identify problems that Alastair Beresford and Sandy Fraser have been talking about all summer at Fraser Research. These guys take a slightly different approach, but adding yet another level of indirection. However, they do agree that the current way that applications use IP addresses mean it is difficult to implement.

The fourth session was on distributed information systems, ie. p2p networks. I didn't pay much attention to this session, so I don't have any comments about the work. Especially, because I don't know what a fluid model is and how it applies to BitTorrent.

The final session called the "Outrageous Opinions" session was quite funny. The idea is that people go up and present fake presentations about something they find. Presentations included, bashing SIGCOMM review committee about rejections, DNS++, bashing SIGCOMM review committee, using slashdot for paper review, bashing SIGCOMM review committee, bashing other papers (this was actually very funny -- there were comments like, we don't need an outrageous opinion session because the papers themselves were very outrageous.) The prize went to someone who got their paper rejected and then wrote to tell about how to dodge the SIGCOMM review committee and how to respond to stupid reviewers.

Anyway, here concludes what I have to say about SIGCOMM for the moment.

You can reply to me about this on Twitter: