Hmm. I am not sure I see a need for collateral here. I basically think of the future rewards as the collateral. In my mind, there should be no outstanding commitment to the network by a neuron’s former owner once the neuron has been transferred.
From a pure trading perspective, I would think that if you stake X ICP for 8 years, and then sell ownership of your neuron at year Y (Y < 8), the total ICP accumulated after the transfer should be equivalent to the amount of ICP you would have received by staking X ICP for N years and then letting it dissolve for the remaining M years, so that N+M=Y and N is maximized. I wonder if there is a way to work backwards from there.
Pricing it this way is clearly not fair to the network because by staking in an 8 year neuron instead of in an N year neuron and never dissolving it, you have obtained additional voting power during those Y years at no additional cost. The amount of extra voting power obtained is essentially equal to:
Extra Voting Power = (8_years_dissolve_delay_bonus - N_years_dissolve_delay_bonus) * (1 + Y_years_age_bonus - N_years_age_bonus) * X * Y
,
ignoring the accumulation of age bonuses. There was also additional maturity earned during those Y years. I am not sure what the exact formula for that is, but I think it can be calculated similarly. Maybe (Extra Voting Power Obtained) + (Extra Maturity Earned)
is a basic starting point for a transfer penalty?
It does become more complicated once you consider compounding of the earned maturity. I think you could still have the payouts though. You would just need enough staked ICP + staked maturity remaining in your neuron to cover the cost of a transfer if you wanted to do that.
Just to circle this back to @bjoernek’s OP:
Phase 1: I support lifting the restrictions on canister controlled neurons as a starting point. I think it presents an interesting new space of possibilities to explore, and from what I am reading here it seems like this does not pose any immediate or likely future threat to the health of the network.
Phase 2 & 3: I think that some form of reduced reward/voting power mechanism for transferable/transfered neurons is interesting. I am not sure I am settled on what exactly is the correct trajectory here. I like the idea of a temporary loss of voting power in the case of a transfer as mentioned by @Seers but I am not sure how to avoid the circumvention of this through a hierarchy of canisters, etc.