96% score on YSlow. Would be 100% if it weren't for uploaded resources (which shouldn't be on a CDN) and the single CSS file (which shouldn't be either, for consistency, and for browser sandbox security permissions).

98% in Google PageSpeed. Remaining suggestions were:
- to minimize HTML (performance bottleneck on the PHP end, not useful due to existing gzip encoding, and risky as can easily break Javascript functionality if the compressor is naive). The Google server-side page speed extension can do this if really desired.
- defer some Javascript loading. As the JS is now merged into a single file to save requests, that means it can't be chopped up -- so this is a trade-off. It'd also be very problematic to do this due to how Composr is put together, as when HTML is parsed much code will assume the JS code for it is already ready, and it'd be confusing to the user for the page to have multiple levels of being loaded.

Now 99% on YSlow. Managed to pipe more through the CDN via some smarter detection and Tempcode flexibility. The remaining 1% is actually hidden inside some factors that rated as 'A', but did not quite get to 100% due to the way the equations were written for them - related to <script> tags.