3) At 140 TPS Bitcoin is still just a corner-case monetary network. It is woefully short of being some sort of global exchange currency. Bumping up to this rate will simply make the solution less defensible for no particular gain.
140TPS is simply the first step to allow Bitcoin to scale.
I think this is the crux if the disagreement.
One side wants Bitcoin to scale. Meanwhile, the other side is saying:
Wait; it does not scale well. Making the block-chain larger will just make it that much harder to hide if governments really crack down.
They have a point: using garlic or onion routing at least triples your bandwidth usage. For small miners running full nodes, their Internet costs will likely exceed their power costs.
I believe that Bitcoin will become much more expensive to use in the future, precisely because it can not scale well. The debate is over how much we can scale the network without awkward centralization problems. I think the recent drop in price has actually promoted more decentralization, based on the
Bitcoin hash-rate distribution (The huge operations can't pay their bills).
The "Gavincoin" advocates have been pointing out that "side-chains" are not really a solution to the scaling problem. It just means the average user has to deal with some kind of centralized exchange. I have noticed that the "MPcoin" proponents don't even try to dispute this: they really don't care as long as Bitcoin remains strong.
I firmly believe that if Bitcoin really takes off, it will mostly replace wire-transfers. VISA and Mastercard will still be common forms of payment. If you want to avoid credit cards for coffee, you will either have to resort to dirty FIAT, or use debit (implying you are still going to need a bank account). However, there is a problem: in it's current form, Bitcoin can not even scale to match wire transfers. I think a 20MB block limit will allow such use, while still making common 100µBTC fees too high for "frivolous" use.
The fees from mining will rise with the price of Bitcoin. Not only will the fees rise in dollar terms, but transaction volume has traditionally gone up every time Bitcoin gains wide attention. In the past there have been (successful) calls to reduce the recommended default fee when the price of Bitcoin rose. I think in the next price rise, we should resist calls to reduce the default fee. Perhaps instead, users can be educated on what a
high-priority transaction is.
For example:
A 1BTC transaction becomes "high priority" within about a day. A 1mBTC transaction becomes "high priority" within about 3 years. The implication being, if you want your low-priority transaction included quickly, you should include the fee (even if it works out to $40 in fiat).
Warning: I have come to the conclusion that we can not simply ignore the "MPcoin" proponents. I also doubt that they will be persuaded a larger block-size is needed until "MPcoin" (forked or not) is run into the ground. One possible alternative is Monero, which automatically scales the block-size based on previous blocks. Monero has a
complicated birth story...involving innovative scamcoins.
Make no mistake: the "Elite" can be effective when they want to be. Their leader, Mircea Popescu, claims to have invented
paywalls now used by the New York Times and other newspapers. Currently, they are in the process of producing a small, cheap "set it and forget it" box that will act as a full node on the "MPcoin" network. They plan to distribute upwards of 10,000 of these. They will be difficult to upgrade to "Gavincoin", and not necessarily due to technical reasons: but also due to inertia. Because they are not designed for easy upgrade-ability, they will likely run modified Bitcoin 0.5.x until they die. They will studiously ignore blocks larger than 1MB.
A competing "Gavincoin" box will need the OS and Bitcoind on separate media from the block-chain storage. The reason is cost: to store 5 years of full 20MB blocks will take about 5.3TB. 5 Years of full "MPcoin" blocks would take only about 262GB (+~30GB for the current chain). The "Gavincoin" solution in version 0.10.x is to start pruning old transactions. Reasonably safe if you go back at least a year (implying 1TB of storage), but is still strictly a reduction in security.
Edit: BTW, full 20MB blocks would have been a problem for my full node recently shut-down. 1MB blocks with 64 connections (+namecoin+P2Pool) were using about 100GB per month; out of a 300GB cap. I am not really aware of consumer-level Internet access that allows 2TB of transfers per month (even if the Head-line "burstable" rate can technically handle it (about 2Mbps down, 4Mbps up)).