Just wanted to leave some general feedback on this article - click here to read it: https://www.theserverside.com/tt/articles/article.tss?l=PipelineVisitor
First, most of the diagrams are too small, I can't read the class labels!
Second, this article appears to have been translated into English, or written in English by a non-English speaker and some of what is written just does not make sense. I'm not knocking non-English speakers but please, can a native speaker proof-read before publishing articles? (this applies regardless of the language(s) published)
The example of exposing part of the application flow for external use is not an example of where this pattern can help, but an example of plain bad design. You decouple internal processes from external access by publishing a distinct set of service end points (web services, RMI or whatever) for external use that are not used internally at all. Then, if changes are required to the internal process you have the choice over whether to leave the service endpoint version unchanged - by retaining and using the old version of the process, or to incorporate the change by using the new version.
The concept of using a collection of name/value pairs to pass parameters is not new - this is simply a variation on the command pattern in miniature, is it not? More an idiom than a pattern. And think of the extra overheads involved in building all those HashMap objects and locating and casting the parameters inside the Visitor, and then performing garbage collection on them.
At the risk of turning into a total critic, the concept of a pipeline is not entirely new either - it simply applies a new term to an existing concept - that of providing a layer of process oriented classes below the presentation tier and above the business tier to orchestrate a business process through multiple user interactions.
The article has some merit but I don't believe it is showing us anything fundamentally new - certainly not fundamental enough to label as a new pattern. It's just a combination of existing, known approaches...