I'd like to start the discussion on the proposal for "defining communicator state with process loss", or recoverability of MPI state in the presence of process loss. To start the discussion I am suggesting that folks have a look at the following paper since I think it relates to quite a number of related topics: Fagg, G., et. al. "Extending the MPI Specification for Process Fault Tolerance on High Performance Computing Systems" http://icl.cs.utk.edu/projectsfiles/ftmpi/pubs/isc2004-FT-MPI.pdf ---- Josh Hursey Graduate Student, Indiana University