Thanks! Yes I believe the SMT implementation as described in the Certificate Transparency paper always does the full 256 hashes per insertion. That seemed unrea

Thanks! Yes I believe the SMT implementation as described in the Certificate Transparency paper always does the full 256 hashes per insertion. That seemed unreasonably high to me too, which is why Quadrable uses a "collapsed leaf" optimisation. The nice thing about this is we only need to do log2(num_records_in_db) hashes on average. If I understand correctly, liburkel does even fewer hashes in the case of common prefixes which is pretty sweet.