“Time spent online” and “bandwidth” are two seperate things.
With DSL and cable, time spent online (where the connection isn’t actually doing anything, just waiting to be used) doesn’t really cost the ISP much. It’s not like an old dial-in service where they had a limited pool of modems and lines–each customer already has a dedicated line anyways. With DSL, your phone line is plugged in to a port in a DSLAM regardless of whether you’re using it. So, with either technology, you can spend as much time as you want online.
When you actually start using the connection though, things change. This is because, even though you may have a 1Mb/s (or whatever) connection, that’s merely the bandwidth between your computer and wherever the ISP starts.
For those who don’t know much about this stuff: when an ISP buys their own connections, which they divide up to their customers, they don’t buy enough bandwidth for all their customers to be using all their bandwidth all the time. In other words, if an ISP has 10,000 customers on 1 Mb/s DSL, they’re not going to buy 10,000 Mb/s worth of bandwidth. This would be horribly expensive, and very wasteful. The reason this is wasteful is because residential customers rarely use all of their connection all the time. Even when you’re “active” online, you’re not using all your connection unless you’re downloading (or uploading) a big file from a server whose own connection is robust enough to saturate yours, or running a server with a lot of people accessing it.
For example, if you’re web browsing, you only use bandwidth when you first access a page: if you go to the Dope, your web browser uses your connection to get a description of what the Dope should look like from the SDMB’s server, and then displays it. You’re not actually using the connection as you read the page, or type a response, or whatever. E-mail works similarly–you’re only using the connection when you click “Send” or when you retreive your new messages.
A good example is how DSLAM’s (the machines that provide DSL service to a line) work. Your line goes to the DSLAM, and it has a speed of, say 768 kb/s. Let’s say 23 other people connect to the same shelf on that DSLAM, and they also have a 768 kb/s rate. Now, even though there 24 768 kb/s lines going into that shelf, typically there’s only going to be two 1,581 kb/s lines going out of the shelf towards the internet (my numbers may be off–it’s been a while since I was directly involved with the hardware). So there’s no way that all customers can use all their lines all the time, because of that bottleneck at the DSLAM. There are further bottlenecks at each step until you get to the “backbone.” If one customer is using his connection in a grossly non-residential manner (constantly downloading and uploading files, with peer-to-peer filesharing, or by running a server, or by working on his midget porn collection 24/7) he’s impacting the service of other customers and costing his ISP money.
300 GB a month is a LOT of data. That’s roughly equivalent to using 2/3’s of a T1 around the clock (a T1 is a type of connection that’s designed and marketed for this kind of high-end constant non-residential use), and a T1 costs a hell of a lot more then a DSL line (at a minimum, around 10 times as much).