i believe there are significant incentives for miner not to create blocks that are TOO big
The relay limitations of all miners are not equal. Orphan risk is unequally distributed based on these limitations, which threatens the viability of groups of miners controlling a minority of hash power = miner centralization risk. See Pieter Wuille's simulation: https://www.reddit.com/r/bitcoin_devlist/comments/3bsvm9/mining_centralization_pressure_from_nonuniform/
What defines "too big?" Without some hard limit that defines this, there is nothing to prevent miners from forking networks.
I see some have little faith in free market here. If miner create too big block, he will find it out after the block become orphaned.
...and? What if this miner found this block first by timestamp, but some miners on the network refuse to build on it, and nodes refuse to relay it, due to its size? If there is no hard consensus rule, what is to stop miners and nodes setting local policies that reject blocks beyond certain criteria? In that case, how will they remain a network?
So rational miner should be conservative with his block size when there is not hard limit.
So what prevent miners from forking networks? The same as today, risk of their blocks orphaned to other longest chain. No need for some hard limit
So what prevent miners from forking networks? The same as today, risk of their blocks orphaned to other longest chain. No need for some hard limit
No, you fundamentally misunderstand how bitcoin works. Orphaning only accounts for the latency issues around relaying blocks to the network. It doesn't address the idea that nodes and miners may refuse to relay blocks beyond a certain size. If so, and some miners continue building on such blocks, we see chain forks. Likely we would have many, many forked ledgers instead of a single global ledger.