For example with an arbitrarily large MAX_BLOCK_SIZE, a mining group hostile to the success of Bitcoin with 1.1% of the network could feasibly solve 1.45 blocks per day on average. If they did so with very large blocks, such that fully 1/3 of their blocks were orphaned and was sending 1GB blocks, this could grow the block chain by 1GB per day.
Absolutely, but scaling the block size according to the projected bandwidth growth rate (based upon the recent bandwidth growth rate) means that 1GB blocks are still many years away, mid-2020s. Storage shows phenomenal potential for further growth and durability.
http://physicsworld.com/cws/article/news/2013/jul/17/5d-superman-memory-crystal-heralds-unlimited-lifetime-data-storageIn ten years time the block reward will be 3.125 btc, and the fees market much healthier, such that a dynamic "soft" limit could be introduced, if that is considered needed. You mention doing the right thing for future generations of Bitcoin users. We all want to do the right thing for them, but we have to concede that they will have far more information to make design and scaling decisions. What is important now is to introduce *some* scalability, ideally so that the ecosystem can grow, as much as computing technology (in the hands of a hobbyist or semi-professional bitcoiner) will allow, in the meantime.
We agree on the need for scaling up. The notion of arbitrarily large sizes was from Gavin's article. He cited a few problems with it, and dismissed them, this was additional to those he mentioned in the article.
In the ideal case, we could devise an automatic and mechanical way to increase MAX_BLOCK_SIZE, so that we are not creating a central authority to decide what the MAX_BLOCK_SIZE ought be, and won't have to rely on their criteria of the day. The information that the future generations of Bitcoin users will have, will come from the block chain, which we also have today, telling us what is needed today. If we are able to solve the problem in a way that accomodates what the block chain is telling us, it may never need to be solved again! It may be able to adjust to changing environment similarly to the way the difficulty does.