It is a matter of perspective. My perspective is that we could show people (among those, people younger than ourselves) who are interested in the data flow paradigm what already exists and how brilliant it is.
The implementor chose NetRexx and not Classic or Object Rexx. I only have spent a lot of nonbillable hours in keeping it working, making sure we have all the source, and lately, that it is usable for every downloader of NetRexx. I know Jeff spent countless hours in making SPEC compatible. The other perspective would be that of the small group that still uses CMS, or has somehow lost access to it. I would not want to develop CMS applications on my workstation, because I like the 3270 CMS interface too much and coiuld not care less about Eclipse and the likes. Give me Lexx or XEDIT. I do however, from time to time, develop small utilities for the Mac’s variant of Unix, or sometimes on Linux in it. Those perform well and, for a change, earn real money - through the time I save compared to doing them any other way. When it generates revenue, I call that production. So from my perspective is it not a toy, but an alternative implementation of something I know and like, and would like others to have the chance of using. If not only for preserving history. I do think that we strayed too far off-topic on a list that is called CMS-PIPELINES. As a last point, ooRexx has some examples that implement pipes in a way that is vaguely reminiscent of Pipelines.. Those are not comparable to Pipes for NetRexx (as it is officially called). I am switching to the NetRexx list for people who are interested in some examples of Pipelines for NetRexx. I do have to test a lot that has changed under the hood, and this is a good opportunity to spend some time on it again. best regards, René Jansen. > On 22 Mar 2019, at 16:07, John P. Hartmann <[email protected]> wrote: > > Right, so NJPipes is a toy that allows you to play around with something that > tastes like what you were used to on your beloved CMS. > > For example, it cannot be used to develop CMS applications on your > workstations. > > Fair enough. > > The regression test is a huge number of pipeline specifications that are > issued from a production level of the pipeline into a test version. Lots of > SPECs stages are used, so if you don't implement SPECs including the 407 > emulator, you won't be able to run it. > > And it has no notion of "reasonably correct". So in summary, it will be of > no use to you. > > On 3/22/19 20:27, René Jansen wrote: >> Hi John, >> that depends entirely on your definition of “reasonable.” Ours might be >> different. >> Mine is: you are able to work “in the vein of” CMS Pipelines by being able >> to compose a commandline consisting of recognisable VM pipeline stages, on >> any computer that runs a JVM. A big difference at the moment is that a >> pipeline in the current implementation is compiled into a (java) .class - >> the NJPipes implementation predated the NetRexx interpreter. (Because in >> contrast to Classic Rexx, NetRexx first had a compiler and after 2000 an >> interpreter.) Also, because of the nature of Unix command lines, it is >> easier to use ! as a stage separator. The impedance mismatch between Java >> and CMS or TSO is certainly there, we can look how fork() would be handled >> for example, but if NetRexx was still supported on CMS by IBM I could show >> you commandlines that are identical. >> NetRexx is a dialect of Rexx, and as such most programs will look different. >> This also goes for non-pipeline filter programs. Readto() and Peekto() are >> there, as is output(). This makes it not too difficult to use ‘more or less’ >> the same Rexx code, and the optimist in me would still call that >> ‘reasonably’ portable. I made my first stages when I only could write Rexx >> and no NetRexx, so it is not too far off. >> You have interested me now in the pipelines regression test; if that runs on >> text data and has no EBCDIC dependencies I am inclined to use that, if I had >> it, to certify the individual stages. I knmow people like Jeff Hennick and >> Ed Tomlinson went to great lengths to be compatible with at least the level >> of Pipelines that they had access to. >> Is this regression test open sourced? Or available somewhere with favorable >> license conditions for the Rexx Language Association? >> best regards, >> René Jansen. >>> On 22 Mar 2019, at 12:54, John P. Hartmann <[email protected]> wrote: >>> >>> How can you say that when it has not passed the pipelines regression test? >>> >>> Can you move at least pipelines REXX filters between VM and your >>> workstation without change? >>> >>> On 3/22/19 17:49, René Jansen wrote: >>>> The implementation (by Ed Tomlinson) follows VM and is reasonably complete.
