Yes, the white paper was still talking in terms of absolute rules (this is why I said that Satoshi did a genius invention, and didn't realize it himself), in terms of what the King decided, and had hence an undefined notion of "honest".  His "to accept alerts from the network nodes if they detect an invalid block" is actually a poor way of killing the consensus decision process by a "vote by majority of nodes" which can easily be Sybil-attacked, and which was the original reason to base consensus upon PoW and not "vote by node" !

Satoshi didn't realize that the immutability and the consensus mechanism he was thinking applied only to the HISTORY of the block chain, was also the mechanism that decided about the protocol.  He thought that he, or his heirs, would remain the central deciders concerning the "laws of the system" (the King and its aristocracy that can dictate the law, say), but that the users of the system including the miners, would be subject to a game-theoretical decentralization, which would keep them on the Nash equilibrium of the rule set the King and his aristocracy decided upon with their de facto software monopoly that contained the Law.  This is why he talked about "honest" nodes, that is to say, nodes that run correctly HIS software with HIS rule set.  So Satoshi was still thinking as himself or his heirs as the central deciders about what is valid and what is invalid, and impose it through their software.

However, the consensus mechanism (the genius invention of Satoshi) applies to EVERYTHING, not only the binary data of the block chain history, but to the protocol too, if the software becomes decentralized too and the central deciders lose their power.

But what you say is evidently not true, in the following sense: suppose that I have 10 times the hash rate of the network under my thumb without anyone knowing, and tomorrow, I fire up my hardware and I orphan the last 100 blocks and replace them by 200 empty blocks (rewinding all the transactions of today).  I can produce 200 empty blocks simply because of my hash rate.  Suppose I keep on mining these blocks for the next 10 weeks.  I will have outpaced all others.  I will have made by far the longest chain with the most PoW.  Difficulty will have gone very high.  All software that is running bitcoin, including all miner software, will mine on top of my blocks after that.  The transactions have been rewound because the only growing block chain is the one which is the "true history" by definition.  The orphaned blocks containing the "real" transactions are now considered non-existing or false blocks.



Of course, but that is exactly because there's a software monopoly (in alt coins even more than in bitcoin).  So evidently, the single software running out there cannot contradict its very own protocol, so with a software monopoly defining the protocol, this situation can of course not happen.

The whole point is that if there is a single block chain out there which has a modified protocol, and if miners all agree amongst themselves to continue mining that chain, using whatever software to do so, then that modified protocol is the true protocol of bitcoin, or of whatever you want to call the coin that goes with it.   If miners are not bound by the central software decider, they are the deciders of the consensus, not only concerning the transaction history, but also of the protocol.

The whole difficulty for miners is to come to any other agreement than the current protocol, because the current protocol is, for them, a Nash equilibrium.  If they know that all their peers are bound by the software monopolist, they can simply follow, and the software monopolist is the true central decider of bitcoin.  But if they are not certain that their peers will run all the same DEVIATION of the actual rule set, then the current status is a Nash equilibrium out of which they cannot escape, unless they collude themselves to decide upon something else.  In that case, their cartel is now the new central decider on bitcoin, because the software monopolist lost its monopoly.  

EDIT: BTW, one should make a distinction between two different notions:

a) hash rate majority
b) hash rate consensus

Hash rate majority can impose history modifications (the famous 51% attack) and protocol modifications that are soft forks onto the entire network.  A typical soft fork imposed upon the network is a blacklist of addresses.  If a majority of hash rate decides to blacklist addresses, and orphan blocks that contain those addresses, then this black list is imposed upon the whole system.

But hash rate majority cannot impose various protocol changes which are hard forks.  This leads us to:

Hash rate consensus.  Hash rate consensus is when all miners accept one-another's consensus decisions, that is when they make one single chain.  Apart from timing problems and accidental orphaning, they accept one-another's blocks and build on top of them.  This hash rate consensus is what I'm talking about when I say that it also determines protocol.

Hash rate consensus can be broken, when a certain part of the miners decide upon one protocol or history, and another part decides upon something else.  Obviously, this can not happen if there is a software monopoly !   But if there is no software monopoly, such consensus breach can happen.  If miners split in two or more groups, and only recognize their peers' blocks as consensual, we simply have a forked chain.

When there is hash rate consensus, the emerging protocol is the "valid" protocol, whatever it is.  If there is no hash rate consensus, there is a fork in the chain, and users can pick one or both of the prongs to do things with, like transact on it, or consider received transactions on it.   When there is hash rate consensus, there only being one block chain out there, there' s nothing else.