This is something that came up in a speculation discussion thread I'm involved in, and I was hoping someone with better knowledge of the technical aspects of bitcoin could clear things up for me.
As I understand it, bitcoin needs to have exponentially greater number of transactions in just a few years in order to keep miners interested as the block reward continues to halve, particularly as the miner fee gets smaller in the face of a rising price of Bitcoin. However, if there is a 1mb block size hard limit, it would seem that something's got to give.
Transaction fees aren't fixed - it's a supply and demand auction for a limited resource. Transactions definitely do not have to increase in number to keep miners interested - it'd be just as easy for fees to increase. An oft-quoted possible future scenario is for Bitcoin to have a trillion dollar market cap, and 0.5% of that market cap going to miners every year in the form of transaction fees - $5 billion a year. That leads to roughly $20/tx, similar to wire orders. (obviously the blockchain will be used for settlement and people's savings in such a scenario) Interestingly, this is almost identical to what transactions currently cost, in terms of inflation. IMO this says a lot about how Bitcoin is used primarily as an investment, rather than as a way to pay people over the internet...
What's good about such a scenario is it's guaranteed that nearly all the money going to transaction fees will go directly to the hashing power that keeps the network secure, and it's guaranteed that mining will be something you can do anonymously and Bitcoin stays decentralized, at the core anyway. You'd need a lot of off-chain stuff, or alt-coins or who knows what else, for low-value transactions, but it's not a model that much different than what people do today. (look at how most Bitcoin users trust central services like blockchain.info and coinbase with their funds %100) If markets demand it, you can get pretty good auditing and fraud prevention and punishment. It's a pretty safe option, but it's not particularly sexy. It's also an option that happens by default: inputs.io, coinbase and others have already popped up with their various solutions and it appears that majority of Bitcoin transactions already happen off-chain anyway. (the figures inputs.io gave me alone a few months ago for example was nearly 50:50 by value, and a clear majority in terms of number of transactions)
One problem with off-chain stuff is it's easy for it to become
too good: if the market finds the centralized and decentralized ways of doing them prove secure enough, that'll undercut mining income by shifting more and more transactions off the blockchain entirely.
You can also raise that limit. The problem here is that the higher the limit is, the more money goes to infrastructure like internet connections, and even worse it makes highly centralized pools more profitable than more decentralized mining. (smaller pools, solo, etc) At the extreme is the frankly insane suggestions to just remove the limit at all and let miners decide, which inevitably lead to dirt-cheap transactions, where the majority of the cost goes to infrastructure rather than hashing power. On the other hand, no-one wants to be the guy deciding the limit - the best we've got is suggestions to have all Bitcoin owners vote on the limit.
But raising definitely has advantages too, and given the inflation reward we'd probably get away with it for another 10 or so years before the inherent economic problems catch up with us.
It's worth remembering that with clever cryptography and some backwards-compatible changes to the core protocol we can definitely raise the limit and still be able to effectively audit the blockchain, at least as long as miners co-operate. We may even be able to retrofit the ability to mine in a decentralized, low-bandwidth, low infrastructure manner, although while that's technically possible, it probably needs
very controversial economic changes to really work properly. Right now the consensus is we need to do something, beyond that though there isn't much agreement. (there is
not consensus to raise the blocksize, at least not in the way the general public thinks, and there also is
not consensus that even if the limit should be raised, raising it will be politically possible)
Anyway, look at it this way: long-term Bitcoin will probably fail eventually due to poorly designed incentives for mining, if not due to some other reason. But another, better designed, crypto-coin can take it's place. In the meantime we can learn a lot from Bitcoin, lessons that'll help design the next generation of crypto-coins.
