Currently, JDBC driver depends on lots of third-party libraries, including Hadoop.
TAJO-910 already showed the possibility to enable TajoMaster to process and just send the result to Client instead of that TajoClient directly reads the result file.
If we use aggressively the above approach, we can allow JDBC and TajoClient to require few third-party libraries. In addition, if we use maven-assembly plugin, like RDBMS, we could make a single file JDBC jar. It will make useability much more higher.