On Mon, 27 December 1999, Eric Germann wrote:

Maybe they could use the marspolarlander.org infrastructure since NASA doesn't have a use for it now ...

At a post-mortum after one of the big events last year, I suggested the federal agencies might want to join together and create a way they could share their web infrastructures for these big events. Different agencies face this problem on a regular basis. Each time they set up a one of a kind project. Afterwards, the vendors pack up all the fancy boxes and go home. If the agencies had a pre-arranged way to distribute information through all the servers they have, they have a lot of server power spread out across many different backbones and could re-use what they've learned from the previous events. On one hand, its good because each project gets the latest, greatest boxes. Government procurement policies aren't designed for the bleeding edge. Allowing vendors to show off their wares, gets a lot of bang for the government buck. On the other hand, it takes six months to get enough network and server power ready. If an agency needs to set up an event site on short notice, they are S.O.L. Since no one has been completely successfull handling event sites, maybe having different vendors experiment with different solutions is best. According to Koskinen the new www.y2k.gov web site can handle an estimated 40 million hits a day. Yes, I know there already was a site using the www.y2k.gov address. It handled about 1 million hits a month. If you remember your performance and simulation class in college, there are all sorts of queues in queuing theory. You get very different results depending on the predicted arrival rates.