From an engineering perspective, I enjoyed the inside look at how Myspace scaled their infrastructure. Scaling is certainly a critical issue for these big sites; one of the reasons that Friendster waned in popularity was that they used algorithms and a back-end infrastructure that were painfully slow for their once-massive user base. <p>On the other hand, a site should be able to handle several hundred thousand users with a single beefy database server, remote static file hosting, and a few load-balance webservers. When you start 'scaling out' to a bigger array of cheap hardware, you introduce complications that make your site harder to maintain and improve. That change will need to happen at some point along the growth curve, but it seems to me that seed stage entrepreneurs need to be most concerned about making a product that people want to use.