There have been some initial and informal discussions about providing a set of APIs for Pig with the goal to separate the "front-end" from the "back-end".
Currently, Pig explicitly utilizes the Hadoop file system and its Map/Reduce query execution engine to produce results. However, some of these functionalities could be made more abstract and encapsulated in a generic back-end, reducing the coupling with the front-end. Olga and Alan suggested me to update the Pig wiki at http://wiki.apache.org/pig/ by adding a new link where I discuss some initial ideas. The wiki http://wiki.apache.org/pig/PigAbstractionLayer proposes some initial thoughts on the abstractions for the Data Storage and Execution Engine for a generic back-end for Pig. The goal of the material is to initiate a discussion on this proposed feature and collect feedback from the Pig community. Thanks, Antonio
