Putting the index on s3

[Brian Fox <br...@reply.infinity.nu>]

Based on my investigations a few weeks ago, it seems that a  
disproportionate amount of traffic to central is for the index. This  
is a good thing since if used properly, people shouldn't have to  
scrape the whole thing. Ahem mvnindex et all...

Anyway, I've setup an account on s3 and will sync the index over to it  
regularly. Once ready, we'll send temp redirects from nginx on central  
to the new location. The bandwidth should be much cheaper and faster  
as a result of the s3 cloud while also improving the access times to  
central for other artifacts.

I figure we'll need to announce the change in parallel because some  
places might have to adjust firewall rules. Does anyone have other  
concerns regarding the change?

--Brian (mobile)