According to @afilini in the developer meeting today, BDK currently uses "dumb" largest first selection. Largest first selection tends to create small input sets in the short term but tends to grind the biggest UTXO to smaller chunks until it eventually causes a huge input set.
As discussed, it would be good to replace it with something that has a better average performance such as oldest first selection or random selection. Optimally, all of the above would be run and the best input set would be chosen per a simple heuristic among the candidate input sets.
This task should coordinate with implementation efforts for #48.
According to @afilini in the developer meeting today, BDK currently uses "dumb" largest first selection. Largest first selection tends to create small input sets in the short term but tends to grind the biggest UTXO to smaller chunks until it eventually causes a huge input set.
As discussed, it would be good to replace it with something that has a better average performance such as oldest first selection or random selection. Optimally, all of the above would be run and the best input set would be chosen per a simple heuristic among the candidate input sets.
This task should coordinate with implementation efforts for #48.