True, in order to provide consistent benefit to users you'd need some sort of mechanism to determine whether it'd be faster to "grab from the swarm" or "grab from the repo".
Perhaps even setting up each already-existing mirror as its own seed-center would help - there are plenty of them around, and by definition they all have the same data. Worst case you end up getting only one connection, and you have the same performance as now, but in the best case you can download from multiple locations at once and get some rudimentary load balancing.
Perhaps even setting up each already-existing mirror as its own seed-center would help - there are plenty of them around, and by definition they all have the same data. Worst case you end up getting only one connection, and you have the same performance as now, but in the best case you can download from multiple locations at once and get some rudimentary load balancing.