The Past

This particular interface had its beginnings at Oregon Graduate Institute circa 1990.  David DiNucci had researched the comonalities between shared memory and message passing interfaces, and had designed an approach called LGDF2 (Large Grain Data Flow 2, named after his advisor's earlier work) to exploit them into a more unified, portable model.  DiNucci implemented LGDF2 (aka F-Nets) on assorted shared memory machines, but putting it onto distributed memory would take more time.  He recognized  its relationship to work done at Caltech on the so-called Crystalline Operating System Reactive Kernel.  Another student at OGI appeared set to take the distributed memory runtime for LGDF2 as a Master's Thesis, but had the sense to instead drop out, start a consulting company, and get rich.  DiNucci finished his PhD on F-Nets without completing the distributed memory runtime.

By 1992, DiNucci was working at NASA Ames Research Center, and had failed to get approval there to develop a successor to LGDF2, called Software Cabling (SC).   The MPI Forum had just formed to develop a standard message passing interface, and it became clear to DiNucci that it would not have the functionality he had envisioned for the distributed memory runtime for LGDF2, so he proposed doing that under the premise that it would fill the gaps in MPI, and NASA accepted and funded its development.  It first began with the name "MPK" (message passing kernel), but that was not descriptive, and it conflicted with an IBM product, so the name was changed to CDS, or more specifically, CDS1.

When the second MPI Forum formed circa 1994, NASA decided there was a high likelihood that MPI2 would finally satisfy any gaps in MPI that were being filled by CDS, so they cancelled CDS funding.  DiNucci then wrote several paper on CDS, and joined the MPI2 Forum One-Sided Communication subcommittee to see if he could facilitate integration of CDS into MPI2, but certain other design constraints in MPI prevented this (despite some support from other members).  He later held a "Birds of a Feather" (BOF) at the SuperComputing '97 conference, where the possible standardization of CDS and/or a similar interface was discussed.  The turnout was small, potentially due to competing BOFs, but those present were (and apparently still are) interested.

Upon leaving NASA in 1998 to form Elepar, DiNucci asked NASA if he could take the (now abandoned) code.  After significant negotiations, he eventually obtained the necessary licenses from NASA (actually the contractor, MRJ), and began redevelopment and extension of CDS at Elepar.  That was the source of this BCR release.