(1) In DumpTypedBytes.java, LoadTypedBytes.java and in HadoopStreaming.java, in printUsage(), "$HADOOP_HOME/hadoop-streaming.jar" is mentioned, which is not the correct path of streaming jar. May be it is better to just say hadoop-streaming.jar instead of giving path.
(2) In DumpTypedBytes.java, in "if (args.length == 0)", before printUsage() call, printing an error message like "too few arguments to dumptb" would be useful.
(3) Would it be better to say in printUsage() in HadoopStreaming.java as "[streamjob]" to signify that streamjob option is optional ?
(4) Description of options/arguments is missing in printUsage() in LoadTypedBytes and DumpTypedBytes. You can give the same description provided printUsage() of HadoopStreaming.java. For dumptb, should we say that the input can be either text file or sequence file ?
Will let you know offline some very minor nits.