Weighting Function ideas

[sternhenri]

Looking over Snow White 4.2 a few things appeared to me:

First, the SPC abstraction is deeply useful to reasoning about EC without letting questions of how stake is determined (in our case stake means power provided to FIL rather than ownership of currency).

I haven't read the Fruitchains paper but there are a few ideas I find very interesting as relates to how we build the power table in SPC:

• Mining Reward goes to the leader, but mining fees goes to fruit (on the separate chain)

This would mean:

• Gossip protocol through which miners exchange recent PoSTs (timestamped for freshness) and post them to secondary chain, for fees
• Fresh PoSTs included in block add to weight (incentive for leader to add to blockchain)
• Block refers to state of fruitchain --> in case of fork, fruits are not lost but can be reincluded in new chain. (Answers Adin's question about cost of forks).

There is also more simply the idea of having blocks include weight, proposed by @nicola of the type:

• "if there are two chains with two different total storage, full nodes should pick the one with higher amount of total storage

This will also add some interesting incentives:

including proofs (since if a miner doesn't include proofs, the total block's power is decreased) including new miners (since if a miner doesn't include new miners, their block's power is less than the others)"

[whyrusleeping]

Potentially interesting work that I did a little while back: https://gist.github.com/whyrusleeping/efcf6041d9f3c3ddddf18df6a1a7f8a4

[sternhenri]

Recognizing your point here about how this is tied to storage power concentration in the network (or lack thereof: a more evenly distributed network will yield faster inclusion times), I am closing this in favor of the current canonical answer:

• 1) The larger issue here is network concentration which will deserve a lot more work thereafter.
• 2) The currently most accessible way to deal with this at the protocol level is by touching up the weight function, but all current enquiries into that lead to a slew of other issues and we don't want to open that can of worms. The weighting function should remain linear in the power of miners as it currently is (and neither superlinear -> centralization, or sublinear -> Bad economic incentives).

For now, miners will rely on fees.