Hi @sipa. As far as I understand, the linearization of a cluster means any algorithm that takes as input a dependency graph of transactions and outputs those transactions in topological order (if B is a child of A, “B->A”, then A must come before B). I understand that filling blocks maximizing fees is a hard problem (literally) but some approximate heuristics can be used to obtain fairly good solutions, for example linearizing the clusters of transactions and merging the resulting sequences. But still it is not clear to me what’s the definition of the optimal linearization. How do you measure the quality of one linearization vs another.
I am asking because I can think of other ways, besides the one described here, to linearize the cluster and I am not able to judge it good or bad because I lack the definition of optimality or the measure of goodness in this problem.