There is an interesting article on CNN about ways to use bandwidth more efficiently using new programs under development -- the ones cited are called Catnap and Naptime.

I thought it was funny that the article pointed out that iPhones go to sleep so often that when they are downloading they keep being moved to the end of the queue, thus slowing them down.

Some of these improvements would require changes in routers, the artiicle says -- maybe this would just be a matter of a firmware upgrade.

The article mentions that some ISPs are putting more stringent limits on downloading, thus causing a problem for relative bandwidth hogs such as smart phones.

Sounds like snake oil to me. : )

If you're moving data by radio then you need power to provide the RF path plus a bunch more for the modulation method that is used. This is required, doesn't matter if you're sending nothing down the line or you've maxed out your available bandwidth - effectively there's no difference between these two states when it comes to power consumption since you're randomizing the data stream anyway.

In the case of wifi there is a lot happening behind the scenes in order to establish the link, you can't tear that down and throw it back up again without introducing a noticeable and significant amount of latency.

For always on data, there's absolutely nothing that can be done other than bettering the state of the art in battery technology, or simply figuring out how to work the sand better such that you have less power draining silicon.

You can minimize power drain for things like MMS or push messaging since these things happen fairly infrequently - it's cheaper (power wise) just to listen for your name and then establish a link, but for average internet use like browsing the web, watching video, music, not going to happen. These activities are noisy and bandwidth intensive.

Just my 2 cents : )

Well, wifi, gsm and 3g don't always have a channel allocated, and don't always have a "link" established in the full sense.

For example, wifi will go to sleep and wake up periodically, listening for the AP's beacons to know if the AP has data waiting for the device, and if so, wakes wifi up to receive that data.

Similar thing for gsm/edge. Re-establishing connection takes a bit of time. You can notice this by idling the connection completely for 10 seconds, then trying a ping. The first ping will sometimes take over 1000msec, but after that it gets better.

3g is the worst for always-on usage. As I've understood it, several devices use the same frequency, and if a device wants to participate, it has to regulate its transmission power and keep transmitting for a minimum period that can, depending on the operator, be quite long. Something like 5 seconds on some operators.
What this means is that one single packet every 5 seconds can cause enormous drain on the battery. A packet every 5 seconds is less than 12 bytes/second, and can use as much power as 1Megabyte/s in the absolute worst case.

Now with that in mind, imagine streaming radio. It's a steady stream, keeping the radio active all the time, consuming enormous amounts of power. How could this be improved? Well, streaming radio is only 128kbit/s, and the N900 3g radio is capable of receive data at 10Mbit/s (in ideal radio conditions). Instead of a constant 128kbit/s stream, you could do a one-second burst at 10Mbit/s, which is good for over a minute of audio. The radio could go back to sleep for 59 seconds until the next burst comes, instead of actively transmitting.

Even further power savings could be done by having all sorts of background apps co-ordinate their activities. For example widgets wanting to update stuff from internet, email updates and so on could agree on doing their network activity at the same time, as to minimize the transmission time of the radio.