Caveat: I am not an RF engineer. I am no longer trusted to do much network config by my team. I run NetBSD rather than a commercial OS and can probably be regarded as a (l)user. We're a mix of platforms, probably more diverse than most smaller, dense deployments. Few of us get to CeBIT or other situations likely to be like this, for worst-case networking. You can't compare this to the density problem of normal offices. I'm not convinced the RF sharing in the A/B/G world works well with older cards, and I am sure we have some people who are using pre-'gold' lucent chipsets locked at 2mbit. I saw a lot of non-printable SSID. I saw a lot of non-IETF, non-hotel SSID. Some I think are in the area, some were undoubtedly the ad-hoc nonsense which crops up. I got better WiFi in the Irish Pub. I am not sure the range of SSID helped. the WEP ssid probably shouldn't be even remotely close to 'ietf62' and having ietf62-wep didn't help me all the time (this is classic 'problem between chair and keyboard' stuff) We have people who freak their power budget. I don't think this is in the wider community interest. We have people who run two cards at once. They probably get a good outcome but I wonder if it helps everyone around them to have two cards cataleptically bouncing between the same channel We have people on laptops with bad engineering. some laptops with built-in WiFi have truly awful issues with large pods of water and fat moving between them and the base-station, some appear to prefer being in specific orientation. I was routinely seeing 8+ base stations on 2 channels. I've said this to people in the NOC, and not yet been contradicted but isn't it meant to be a bad idea to have that many base stations visible on the same channel? I thought even spread spectrum had limits to how much RF you could use with other people. The reports that turning off one of A or G improved things makes me think this has a lot to do with the link layer, the RF. the DHCP service was odd. If the net is not actually congested, then why would a DHCP server take so long to serve? I don't see why the beast was stuck in backoff land for so long. Of course, if it too was seeing melt-down of the carrier, the ARP must have been lost along with all the good bits, but when the net was palpably 'up' -Why does a DHCP server struggle? If its not scaling for a large flat net, should we be walking back to smaller partitions, or to some other address discovery service which does scale to large flat nets? Not all Mac users had a good time. This is often a clue that things are bad in the infrastructure. I don't have one, and they love to tell me how perfect it was for them on their MacOS, but they just weren't doing it, and I think that also says there was something wrong out there on the aether. I respect the effort the NOC put in. I know its hard to do this stuff, I do it writ small on far smaller conferences, and I know how hard it can be to make it fly when the users are grumpy. But we're not going to get to a better place without recognizing this was NOT a good network, even relatively speaking inside the history of IETF nets. -George _______________________________________________ Ietf@xxxxxxxx https://www1.ietf.org/mailman/listinfo/ietf