According to CASSANDRA-3134, the preferred mechanism for binary I/O with streaming is Typed bytes. Two streaming based libraries (Hadoopy and dumbo, both for python) use typedbytes for other types of binary input, creating some support. So if we just implemented an R reader/writer for typed bytes and CASSANDRA-3134 is fixed, then we should have cassandra support and also some other binary i/O options should open up. See also HADOOP-1722, the original issue about typedbytes.
Will it include an R client for http://thrift.apache.org/ used by cassandra?
see also CASSANDRA-3134