can you elaborate on these design decisions (they add extra complexity):
- why partition along rows and columns, instead of just partitioning across rows?
- why allow lumpy custody sizes, instead of just making a uniform custody chunk size? super-full nodes could just be multiple peers in the network as in beacon
a couple other questions:
- is peerdas data intended to be ephemeral as in 4844 or more permanent?
- what are the desired timing properties of peerdas? live available data every single block?