Well, all the currently proposed volume-based fees are strictly greater than the space-based fee because they are charged per output and no output is >1KB. I’m proposing we charge the volume based fee for each individual non-change output and the space based fee if any change outputs are present. A chunking event is a transaction where every output is a change output. A change output is determined by whether or not the output volume for an address (sum of all outputs for the address) is less than the input volume for that address, and the input volume for that address is non-zero (i.e. that address signed the Txn).
A change address when not in avatar mode is of course treated as a non-change address.
Also, the chunking should be a fibonnacci chunking.
Dividing a 100 NBT output would take 0.2 KB for the input +0.04/output across 9 outputs so the whole transaction would be under 1 KB and would cost just the minimum change fee (0.01 NBT). Then, buying that 25 NBT cab ride we would use the 1,3, and 21 NBT outputs again remaining under 1 KB and paying just the volume-based fee on the 25 NBT (0.095 NBT).
Doing this in avatar mode would entail paying the volume-based fee on the 25 NBT and then 0.01 NBT additionally for the space (up to 1 KB) because of the change output. The result is actually that you end up paying the same as without avatar mode because either you are paying for a chunking event or change. Of course, you could also chunk your change output into fibonacci chunks, thus filling up the rest of the 1 KB that you bought and avoiding having to pay the space fee on the next transaction (Txns where all outputs are unique from all inputs do not have to pay the space fee because the volume fee is more than that anyway).
The underlying premise here relies on the concept that all inputs are signed and thus if an output goes to the same address as an input but with less volume, it is by definition a change address (in avatar mode).
