That is a good point. I have been trying to make something that will scale to 500,000 users making 10 posts a day, but so far I have no users and they are making no posts a day. Even assuming ten thousand users making ten posts a day, things become much more realistic in the bandwidth department. Even if we assume all messages are 10KB, and the metadata associated with a message is 1KB: 10,000 users making 10 posts a day brings metadata to about 98 megabytes per day * 10,000 users = about 960 gigabytes in metadata bandwidth per day (98 megabytes per user per day). 10,000 users * 5,000 messages per day * 10 KB per message = 480 gigabytes in message data for receive assuming Pynchon Gate PIR, there are 100,000 messages per day means each query is 100,000 bits about 12.2 KB * 5,000 posts * 10,000 users = 572 gigabytes in request data 960 + 572 + 480 = 2,012 / 1,024 = 1.96 terabytes of bandwidth required per day, which is realistically affordable with a high end server package. On the other hand, assuming 500,000 users making 10 posts a day brings metadata to 4.8 gigabytes * 500,000 = 2343 terabytes in metadata a day, which is completely unrealistic. But I would much rather make something that can scale to 500,000 users than something that only scales to 10,000 users.