Pipeline tool running on IBM's CMS and TSO operating systems, developed by JohnHartmann? of IBM Denmark.
CMS/TSO Pipelines implemented some of the concepts of UNIX on top of CMS (and more recently TSO), plus a network topology similar to that of FlowBasedProgramming (FBP). It also runs on zOS.
CMS/TSO Pipelines is said to maintain record semantics, which means that multi-stream pipelines can be made "predictable". This has to do with the concept of 'record delay' - CMS/TSO Pipelines wants to assure that records taking different paths through a network will not get out of step with each other, as long as the network only contains stages that do not "delay the record". This is different from the philosophy of FBP, which doesn't worry about keeping data items in different paths in step, thus allowing a much looser, and more distributable, structure. In FBP we realize that it is expensive to resynchronize data, so we recommend that, if synchronism is important (and it usually isn't), it is better to keep the data items (IPs) in step all through the processing. There are several ways of doing this - one example is TreeIPs. Of course, this can also result in a less efficient use of resources. --PaulMorrison
Visit http://vm.marist.edu/~pipeline for the current status of CMS/TSO Pipelines.