i) Is it possible to solve the problem of lagged price feeds as follows?
Settlement/trade is effective instantly, but the settlement value at that timestamp is determined in a verifiable manner after the fact according to a prescribed algorithm/formula. So for instance, it could be based on a formula using a specified weighted average of mid-prices from a given set of exchanges. The "formula" could be voted in by delegates, and changed in future if necessary according to another vote. [Edit] Instantaneous price feeds by delegates would be merely indicative - the true settlement value is deterministic afterward but still requires way to form a consensus on value at past timestamps. [end edit].
Such an approach is similar to how many derivatives are traditionally settled.
The advantage of this approach is it removes the need for a 24 hour delay before becoming effective, it merely requires some delay before a value transfer occurs in line with the calculated settlement.
This does not remove the need for queuing discussed further below.
ii) Creation and cancellation can both occur at the price feed
2. Settlement at 1% of the feed price creates a "liquidity" imbalance where you can essentially sell a large volume of USD without bidding up the internal market. Shorts must push the USD value down to acquire a large position, but longs are not forced to push it up to settle a large position.
In principle, if there is comfort with the settlement procedure, there is no reason why new currency creation cannot also be forced to occur at the price feed, just like currency cancellation. Where there are offsetting requests for creation and cancellation, these help fulfil the needs of both sides more quickly.
You can consider this to be a creation/cancellation market. A completely free market in bitUSD versus BTS and other assets would exist outside it, and allow for more urgent exchange when the creation/cancellation market is subject to queues.
In fact the Currency Creation Market described in my whitepaper gives the most generalised form of this, inviting any parties that wish to transact at the price feed.
iii) Queuing is probably a reasonable idea, but results from insufficient liquidity in the collateral
So without further ado I would like to suggest a compromise that should balance everything out nicely.
1) Limit the amount of USD that can be force-settled each day to 1% of the supply. This would take it almost a year if there were constant redemptions to free the entire supply.
2) When a user requests redemption they are placed in a queue that is filled in the order of redemption with at least a 24 hour delay.
There needs to be a way to deal with potential manipulation of the BTS price. My initial feeling is that the approach of queuing is a workable solution, although the 1% should be faster and vary with liquidity in the external BTS market. Having said that, I do think we need to consider the wider ramifications it might have on the design of a bond market, so we see the big picture before implementing it.
The implication is that the price of the bitUSD will float away from the peg in external markets, reflecting the cost of time in the queues. However as BM mentioned, such spreads would probably occur anyway because of the uncertainty in converting large settlement parcels of BTS into real USD.
The root problem is that we are using a form of financial collateral (BTS) that is not very liquid, unlike say government bonds, a key form of collateral in the traditional financial world. We don't have any other option right now, although for those curious I have suggested elsewhere that it might be possible to create a bitBTC collateral backed by real BTC, that might be more liquid and less easily manipulated, which could improve market confidence.
iv) We should explore how to incentivise shorts when they are unwilling to take positions
I think we need to consider this further. Although BitAsset 3.0 does not require shorts to pay yield, it is just as restrained at the zero interest bound as any yield based approach would be. It may be possible for example to have a completely flexible yield not subject to the zero interest bound if we create an enforceable payment mechanism from longs to shorts. I've considered this separately in the form of deposit accounts, but it needs more work.