Our current indexing implementation allows for string data only to be written to the index. In order to take advantage of some more compressed and more performant data formats, we would need to allow binary data to be written. Note that supporting binary data directly would be a large change. Alternately, we could convert the binary data to an ASCII format, like base64, but in my opinion, the additional overhead of performing that conversion may defeat the benefit we would gain from using one of the binary formats directly. Another alternate might be to use a system next to indexing. That could be a separate file (or files) that contain the binary data where the entries in the index simply point to those files.