[loewenskind]

Bad analogy. If you waste 50,000 gallons of water by leaving your water hose on while you're on vacation, that's 50k gallons that the community no longer has [1]. The only way someone downloading e.g. 5gig of data in a month can cause anyone else to not be able to download is if the line is running at 100% capacity 24/7. The only possible impact the person could have is making communications take a little longer during the times he's bursting. But this is easily fixed by bandwidth caps and burst limits.

Do you have any evidence that lines are getting anywhere close to 100% utilization perpetually?

[1] I'm simplifying of course, but this is the basic idea.

[jsz0]

On big access networks you have to stay far away from 100% capacity. 60-75% is the minimum where you will start seeing the negative effects of channel contention. On a cellular network sharing data & voice you also have a good amount of bandwidth reserved for real time services (phone calls) to factor in along with all the PHY/MAC overhead. There's probably some QoS overhead to calculate in also. I'd say there's a very good chance AT&T is running into channel capacity problems (further evidence being dropped calls/failed calls when devices have good signal. That means they're requesting a real time service and there are simply not bits left over to grant it. Try again in 5 seconds and it works because bandwidth has been allocated/reserved for voice calls)

[loewenskind]

If they ever drop calls from capacity problems then something has gone horribly wrong (either tech wise or provisioning).

But my point was that bandwidth is (or should be) renewable. If you do a download when everyone else is doing one your traffic should be getting shaped down to the provisioned minimum until there is room for it. This will make the traffic slower during those times, but there are going to be times where nearly no traffic is on the line.