"Protocols are for when there's no good centralized way of getting things done" is actually not as crazy statement as you think, since it also means "If there's a good decentralized of way of doing something via protocols, you don't need a centralized approach".
What I meant by "getting things done" was efficiency. Decentralized and centralized systems each have their pros and cons, and efficiency is definitely not one of the pros of decentralized systems.
I do apologize for the incoherency in the post but I still don't think I didn't make any incorrect points. I think it would be if you could provide a counter example to my statements.
The economy is a simple example of a system where decentralized == efficiency and centralized == inefficiency.
The article's author takes the single example of email, and, extrapolating from that, assumes that it's impossible to build low-latency decentralized publishing protocols.
And his statement, "Protocols are for when there’s no good centralized way of getting things done," is so ridiculous, I don't even know where to begin.