A new study shows that absolute decentralization makes scaling impossible
Karov and I have been having this ongoing debate. This news ties directly into the ongoing debate we have had for many months. It's proven fact that to go mainstream requires high transactions per second (scaling) but the research shows it's impossible to scale without sacrificing some decentralization to do it. This is what I said the whole time which is why I think decentralization if it gets in the way of scalability more than it provides additional security is in essence a waste.
So this proves what I said all along. We cannot have absolute decentralization and mainstreamability. We can have enough decentralization to preserve security though. My view is decentralization is a feature with a purpose (promote security). When security is promoted then to increase the level of decentralization does not further increase security. So you end up over paying for security and reducing mainstreamability if you get it wrong.
Vitalik voiced his opinion on this but where I disagree is where he puts "security" in it's own league separate from decentralization. In my view decentralization only has value because it promotes security. It does not promote efficiency. It does not mean anything to have decentralization for the sake of itself. So what does decentralization exist for? Well it always existed to promote security and resilience of the network. For example the Internet was supposed to survive a nuclear attack so that communications could be maintained. This puts context and reason behind why decentralization matters rather than mere decentralism which is like "gospel truth".
What I want is to see these life changing technologies as accessible as possible. This requires putting an emphasis on mainstreamability. This then requires putting an emphasis on scalability. That requires striking a balance between security and scalability. If security is not at risk then more decentralization isn't going to make the network more secure and will only make the network more expensive to maintain (more cost per unit of security gained). If we measure security by the statistics and incidents then we can come up with a risk spectrum and find out whether or not spending more on decentralization is actually improving security or not.
Do I think more decentralization is better if scalability is the same? Of course! Do I think more decentralization is better if scalability is sacrificed and security is the same as before? NO! DPOS is sufficiently decentralized to run Steem which proves in practice (not theory) that the security is sufficient. It is true that we could improve upon DPOS and I actually do not find EOS to be secure or decentralized enough based on how easy it was to suspend the entire network politically but I also do not think the ideal is to have a network so decentralized that a community cannot find a way to come together to suspend a network. So while I do not think it should be easy I also do not think the ideal is to have it be impossible. It should be sufficiently hard politically, socially, technically, that it does not happen frequently.