This is a very interesting article, as I'd heard references to SPDY before, but didn't know what it was.

However, I am bothered by the fact that this is a piece on performance yet it contains no hard data. Without quantifying the scale of the improvement, I have no idea if this is a 2% optimization, or a 15% one.

Also, does it matter if only one side of a connection has slow start enabled? IE, does it require symmetrical SPDY, or would I see improvements on my server and desktop if I made this change?

(Side note: the requirement for performance data is analogous to ROI data in business.)