2/2 Put differently, is there reasons to not make available high computational cost functions as precompiles as long as we price them efficiently? Or, are we risking increasing the specs for nodes too much if we start adding these? Maybe Iām not fully grasping the role of precompiles.
suspect that its opens multiple attack surfaces which are really hard to solve generally
It's more a question of adoption and support. For instance, BLS curves have been on the shortlist for a while, but still aren't available. Each crypto primitive we make available in the evm (regardless if via opcode or precompile) increases a support and testing radius, vs. demand.
It sounds redundant to add Ed25519 signatures when you already have ECDSA signatures. Shor's algorithm can compute discrete logs in elliptic curves and thereby recover the secret scalar from a public Ed25519 key, which you can use to forge signatures of your choice. Better use something quantum resistant like SPHINCS+