[loewenskind]

You may be right about the result but the gas analogy doesn't work. To make it work you would have to image customers all getting gas from the same tank. While they're filling their vehicles the tank is getting lower but the instant they stop the tank is full again. Not slowly filling up with gas, instantly full capacity again.

[batiudrami]

I'm not sure what difference this makes. I understand that bandwidth is limited by transfer rate and not data transferred, but there is still a limit which the company does not wish to reach.

Even if such a fuel tank existed, would it be unreasonable for BP to charge by the litre of fuel removed? I think it'd be the most sensible way to avoid demand overriding supply, rather than charging everyone increasing amounts for access to the fuel source.

[loewenskind]

>Even if such a fuel tank existed, would it be unreasonable for BP to charge by the litre of fuel removed?

But effectively no fuel was removed. Look in the tank, it's full right? What you would charge for is accessing the tank at all because unless it is sitting at 0 constantly (complete continuous usage) there is effectively no difference between someone who took 1 gallon and someone who took 1,000 gallons.

EDIT: Your point "but there is still a limit which the company does not wish to reach" is certainly valid, but properly dealing with this is complex. Simply capping to e.g. 5 gig is just going to leave you with empty lines for the last part of every month. Since your problem is utilization, not capacity this doesn't even address it. You had room for twice as much access but instead you look slow and expensive.

If you had a company large enough to run its own networks how would you deal with growing capacity? By putting some kind of bizarre transfer cap? No, that would just get you fired. Instead, you would shape traffic to even out the utilization and add infra as soon as you passed some utilization threshold.

[batiudrami]

The way this is combated by Australian ISPs is to offer 'off peak quota' - a kind of bonus for when bandwidth utilisation is low, as well as having people's quotas reset at different times of month. This isn't a great solution, and would not work for phones as the off peak quota is used for things such as torrenting, when people aren't at their computers.

The problem with shaping traffic is that it means companies simply shape during busy periods, rather than increasing capacity. By offering download limits at set speeds, ISPs guarantee a speed you will receive (limited only by the quality of copper in your phone line).

My point is largely that charging people who use download the most more money (and therefore, on average, are increasing the load at peak usage most often) is the most sensible method of controlling load. If your experience with mobile internet is anything like mine, consumers certainly do not want bandwidth shaping making it any slower.

[loewenskind]

Shaping is only part of the solution. Normal utilization will consist of spikes. Shaping is just to smooth those spikes out as much as possible. When there is no utilization then you can use up the whole pipe if you like. If utilization is consistently passed some threshold then providers should absolutely increase their infra.

That said, the Australian solution sounds quite good as far as metering goes.