A proposal for a demand-based block size adjustment

A proposal for a demand-based block size adjustment aimed at reducing fees during peak hours and decreasing them during off-peak hours, as well as gracefully handling exogenous shocks in transaction volume

At present, Monero has a dynamic block size system which works as such (copied from StackExchange):

1. Determine max block size as min(trimmed_avg(last 100 blocks), 60 KiB)
2. If the block size is above the max block size, apply a penalty.

Then, if miners are rational, they should always add transactions as long as they stand to gain from it. The marginal cost of adding a transaction to a non-full block is zero. The marginal cost of adding a transaction to a full block is given by the derivative of the penalty function. Since it is quadratic, the marginal cost should be linear.

This ensures that there can't be any congestion, as in other cryptocurrencies. On the other hand, miners have no incentives to make the blocks smaller than the limit. Indeed, they have an incentive not to, since small blocks would enable them to charge less rent.†

I propose that this algorithm should be changed slightly:

1. Determine max block size as min(trimmed_avg(last 100 blocks), 60 KiB) + burnt_scrip
2. If the block size is above the max block size, apply a penalty.
3. If the block size is below the max block size, give the miner (max_size - block_size) units of scrip.

For example, if the max size is 1 MiB, and the miner only used 700 KiB, the miner should get 324 units of scrip.

The miner can then use this scrip later on to increase the block size temporarily, as shown in step 1.

This means that miners would have a good incentive to create blocks that are not full, keeping the size of the chain somewhat lower.

Notes:

* I expect that miners will just sell the scrip on the open market as soon as they can. This is a slight incentive to pooled mining.

* The scrip would, I imagine, be implemented as some kind of colored coin.

* It may be a prudent idea to make the scrip become useless after, let's say, a year, to prevent hoarding and other strange behavior, as well as to enable the idea to be gracefully retired if it turns out to be useless.

† It's the case that low-fee transactions won't be relayed, but as the purpose is just to fill up the blocks, nothing prevents the miner from simply creating lots of garbage transactions and paying himself the fee.