Would there be any merit to an automatic discovery and download? Or would it be better to rely on communication among users about the availability of other nodes on the same network? An example I see is a college network, where nodes are likely to exist but be unknown, as opposed to a home network, where a user will know if they are running Bitcoin on another computer. – Colin Dean – 2012-11-20T17:32:52.120

1Bitcoin is a financial app (i.e., money is involved. As such, you don't cut corners.) A node is best secured when run in a manner in which it gets blocks from randomly selected peers. This approach does cause performance that is less than you can get if you manually connect to other local nodes but you would only do so if you trusted those nodes. In a college environment, you would not want to trust getting all your blocks from peers within that network. Again, security is of greater importance than efficient use of bandwidth. – Stephen Gornick – 2012-11-23T08:13:16.340

That's an excellent point that I had not considered, that security trumps efficiency. However, given the nature of the blockchain, how difficult would it be for a malicious peer to distribute a fake blockchain if our client would eventually contact the outside world for updates, or even an initial portion of the blockchain? – Colin Dean – 2012-11-23T22:04:44.767