Also, I didn't realise the ma block size can be changed by the user of each node/core client. But if you're going to double the limit, why have a limit at all? It's not likely to be over 1GB in size so set that as the maximum and then everyone gets fast transactions, although then the network has to be changed with the difficulty based on different size blocks (doesn't it)?
it doesnt work like that.
imagine there are 5 people.
A says he can read one page of a book in a minute.
B says he can read 2 pages of a book in a minute.
C says he can read 16 pages of a book in a minute.
D says he can read 3 pages of a book in a minute.
E says he can read 5 pages of a book in a minute.
ok so. someone wants to ensure a fair system and although only person A is holding up the group a 20% risk of system error is risky. so only one page of a book is handed out in a minute to ensure everyone can handle whats been handed out.
later person A reveals he has lied to the group and can handle 2 pages a minute. so now 2 pages a minute can be handed out a minute because 100% of the group can handle it.
after a couple years person A learns to read a bit faster.. and 3 pages are handed out because the whole group can handle it..
so although the group can individually have their own preferences the person making the pages (blocks) only makes it at a rate the whole group as a whole can handle. which is usually help back by the ones with the lowest number..
core are holding it back.
just ask yourself what would happen if in 2013, core decided to have set the block limit at 0.5mb due to sipa's accidental bug when he swapped database types and cause a bit of orphan drama.