Quote from: Technomage on February 21, 2013, 09:17:48 PM

Whatever happens, we must strive to keep most transactions in the blockchain (most, not necessarily all), otherwise the whole idea behind Bitcoin is more or less doomed.

I disagree with this assumption. It is entirely within reason for 'bitcoin clearinghouses' to exist that process the majority of out-of-band transactions, while remaining functionally decentralized. The most important part is that it remains forever possible for anyone with the resources to jump onto the network and compete. It does not have to be a cheap endeavor.

That said, I would say that it would be wise endeavor that the main network is capable of processing enough transactions that it's unlikley that any single out-of-band processor can capture more than 50% of all transactions. Practially, I don't think that this is a real worry unless Bitcoin never really becomes a global currency; for it's unlikely that users in China or Russia are going to favor a processing agency in the United States for their mostly local and mostly daily purchasing needs. Thus, whichever processor grows up to be the Bitcoin version of Paypal in the United States is at a disadvantage towards capturing the Chinese or Russian markets due to cultural differences and simply mistrust of Americans.

In this context, such distrust is both well founded and advantageous to limiting centralization.

Furthermore, it's important that competition to any of the market dominators in the out-of-band transaction processing never be 'locked-out' from access to the main bitcoin chain nor (hopefully) co-opted by government policy. The real problem is that we can't predict how these things are going to go with any real certainty. What we can predict is how large of a block must be possible to sustain a particular transaction rate at the current average transaction size. Personally, I consider the rate that Paypal can process transactions to be a practical minimum for consideration of a hard code fork, so nothing less than a 10MB hard limit is worth the trouble. Fortunately, we are still years away from this problem, and have plenty of time to hash out the details. In the meantime, we can just agree to raise the soft limit, or agree to simply let the rest of the miners to comment it out at their own desires. Even the occasional full block isn't really a problem, so long as we don't have full blocks for hours on end.