09-09-2013, 02:13 AM
I think I've fixed a major issue with the client consuming too much RAM when connected to MCS. We didn't send any UnloadChunk packets for 1.3.2+ clients - because the protocol didn't have an UnloadChunk packet. What remained hidden was that chunks were unloaded using the regular ChunkData packet. I've implemented that now, so the clients should be much more manageable RAM-wise.