group.index // txt file poco propertie format group hash = folder name (group_1 -> ...) uint32 group_folder // binary, konto hash = kto index uint32 (first in transaction, getting 1), sorted by hash // hash at end of file // sort in memory, file is unordered, maybe for small files linear search is faster // put last used in cache pubkeys_ _.index // contain group states // like last kto index // using leveldb .state // lowest kto index, highest kto index // kto index = count, [file positions where transaction which contain kto starts] // sorted // write out after block finished, until than exist only in memory // hash at end blk0000000X.index // size, serialized transaction [loop] // hash at end blk0000000X.data