I used one in a bittorrent-like protocol, each block hash was added to the filter then sent to peers. Peers could then have a high accuracy way of checking if a peer has a block without having to communicate with them.
That's really quite cool, how many blocks could an instance potentially have though? Seems unlikely you'd be saving that much space over having a list
Sorry I wasn't clear, each peer maintains their own personal bloom filter that contains the hashes of the blocks that they have. The filters are sent upon connections with another peer and updated every so often.
Is there any code that you can point to with this implementation?
Sorry I haven't released anything, potentially in the future.