Larger block sizes don't scale the network, they merely increase how much load we allow the network to bear.
Very well put, Jameson. And the cost of bearing this load must be paid for. And unless we’re willing to accept that computational resources are finite and subject to the same economic issues as any other finite resource, our incentive model collapses the security of the network will be significantly at risk. Whatever your usability concerns may be regarding fees, when the security model’s busted usability issues are moot.
Larger blocks support more transactions…but they also incur Ω(n) overhead in bandwidth, CPU, and space. These are finite resources that must be paid for somehow…and as we all already know miners are willing to cut corners on all this and push the costs onto others (not to mention wallets and online block explorers). And who can really blame them? It’s rational behavior given the skewed incentives.
On the flip side, the scalability proposals will still require larger blocks if we are ever to support anything close to resembling "mainstream" usage. This is not an either/or proposition - we clearly need both.
Mainstream usage of cryptocurrency will be enabled primarily by direct party-to-party contract negotiation…with the use of the blockchain primarily as a dispute resolution mechanism. The block size isn’t about scaling but about supply and demand of finite resources. As demand for block space increases, we can address it either by increasing computational resources (block size) or by increasing fees. But to do the former we need a way to offset the increase in cost by making sure that those who contribute said resources have incentive to do so.