teryret

It also backed by some sound economics (the Nobel Prize winning Coase theorem) and one of our advisors, Alex Tabarrok, an econ chair at GMU, thinks it is the best possible design.

Setting aside the selection bias (of course your advisor thinks its a good idea, if he thought you were doing the wrong thing he wouldn't be your advisor). The Coase theorem just means that with low transaction costs you reach a Pareto optimal state. The problem is, Pareto optimality is a local optimum, so all you get from the Coase theorem is the argument that "the system will reach a reasonably okay solution". What you do not get is any reason to suspect that the solution is anywhere near the "best possible" anything (that'd be the global optimum).

And in general you wouldn't expect it either. Everything in economics is based on self-interested parties. So as a thought experiment I put to you, how could it be that a collection of self-interested parties could ever out preform a collection of cooperative parties (which in the case of distributed protocols you can have)? Obviously it cannot; the proof is trivial, cooperative parties can act as self-interested parties, but not vice versa, so the algorithms available based on cooperation are a strict superset of the algorithms available to economics. That's why economics people worry about local maxima but computer scientists don't.

Sure. I'm basically pointing out the flaws in the logic of the marketing line. In optimization (real optimization, the computer science kind) there are two kinds of optima, local and global. Global optima are solutions that cannot be improved on (Helen of Troy), whereas local optima are solutions that are not adjacent to better solutions (the most beautiful girl in the whole wide room). Econ people like to bandy about the term "Pareto optimal" because it has the word "optimal" in it which suggests it's the best possible. Unfortunately when you look at the definition of Pareto optimality you find that it is really just another name for a local optimum.

In terms of LBRY, what I'm suggesting is that they look beyond the econ literature to find name distribution algorithms that may work better than "constantly auctioning". For example, they might adapt a distributed trust algorithm like a variant on the consensus algorithm used by the Ripple cryptocurrency (which flopped hard as a currency, but consensus was actually brilliant).

The details are at https://ripple.com/knowledge_center/the-ripple-ledger-consensus-process/

but the TLDR is that Who Owns What is recorded in a distributed database (ie who owns how much Ripple, or who has what domain name, etc), and writing new values to that database (ie changing ownership of something) requires a large number of opposed parties to agree that a transaction obeys the rules (and isn't fraudulent).

Such a mechanism could probably be adapted to work for LBRY, I'll put some thought into this tonight and post back if I come up with it. It will almost certainly involve a collective human effort to mediate conflicts (not entirely dissimilar to voting here on reddit)

I mean, that's basically what's going on right now with patents... anything worth owning is more accessible to corporations than individuals. And thus, individuals will never have anything worth owning.

And this is why you should never let economists design anything that end users actually have to interact with.

FTFY