Thanks Garrett.
So increases in the difficulty level also increases the amount of time it takes to compute a block? Why don't we compute blocks as quickly as possible, then use a lottery system to award the 50 BTC "prize" to the first hasher of a random block in a 10 minute time period? Seems like we're just wasting computing power that could be put to better use...
Unless the goal is to artificially limit the number of transactions that can take place in 10 minutes, to force people to eventually pay transaction fees.
But if that's the goal, what happens when adoption of bitcoins is 100x what it is at now? If there can still only be X # of transactions per block, and the difficulty adjusts itself to where only one block is computed every 10 minutes, how in the world will the system be able to handle an increase in usage? Transaction fees would have to increase to the point of fewer people wanting to use the system, which would be the opposite of what we all want, right?
Difficulty only increases when we're finding blocks, on average, more than every 10 minutes. The 10 minute goal was just a best guess by Satoshi for how long it would take transactions to properly propagate, and to give people with slower connections a chance to help, and at the same time avoid chain splits.
Like was mentioned above, block size can be increased, it's only this small to avoid spam, etc. Already there has been an increase in the number of transactions lately, even with the pools now using sendmany rather than just send.
The blocks are created with the transactions because of the proof of work that the block is - it's hard to compute that stuff, look at how much computing power we're throwing at hashing out a sufficiently small number, and we get one every 7 minutes or so (difficulty is gonna go way up!). So if we just made a block every 10 minutes or every 10 seconds, there'd be no proof of work there, and double spending would be possible.