Currently we hold the entire archive in Redis. This means we can't support more posts on the board than we have RAM for.

The fix for this is to generate JSON and HTML serialization of /archive/ threads. The only real design question left is how to handle >>links to archived threads; should we keep the post index in memory for now? If so, how do we load it? If not, >>link-loading will have to become asynchronous, and that will be super-messy...