Thanks Berry, Kris, Lionel, and those who responded privately.
More implementations than I realized. Wow!

How do you explain the difference between Hartmann pipes and
Thompson/McIlroy pipes?
It's like a baker who makes donuts, cakes, and bread. With
McIlroy/Thompson pipes, the donuts come out one at a time. Fine. But
most people purchase box (or a bag), which equates to a *record* in
Hartmann pipes. Then the cakes and bread come out via different streams.

Trying to be somewhat of a peacemaker, one observes that one-at-a-time
is what happens under the covers. (Okay, real donuts DO tend to be
cooked in batches. But let me continue the analogy.) The problem then is
handling of the sales ... natural boundaries. I guess I'm saying that
Unix pipes have their place. We just want more.

To establish flow control, it occurred to me that the consumer can send
*upstream* inquiring of the producer how many bytes are coming and
signalling when it is ready to receive. So using a pair of
Thompson/McIlroy pipes we can effect a Hartmann connector. This works.
It's not CMS Pipelines. It's merely another attempt to mimic the behavior.

My goal was to avoid the JVM. I also don't have a Swift compiler (to try
one of the Craig Edwards projects). And fond as I am of Rexx, I didn't
want a dependency on it for the underlying functionality. I can
currently write stages in C, Rexx, or COBOL (no joke). I won't be able
to complete the JNI for Java by time of preso at the workshop. Oh well.

Meanwhile, I continue to dig into these older implementations and learn
as much as I can from them.

-- R; <><


On 6/9/24 7:43 PM, Rick Troth wrote:
hello gang --

I've been trying to gather a list of pipelines implementations apart
from CMS/TSO.
Specifically, I know about NetRexx Java Pipes, and I think there was
another Java implementation. I also found an implementation in Swift.
Are there others? This is my current question.

I'm scheduled to present at the VM Workshop on "Pervasive Plumbing -
Pipelines for Everyone". My own attempt has finally borne fruit. So
many stages to write. So little time. I want to make sure that the
talk covers sufficient ground and gives credit to the other developers.

thanks!



--
-- R; <><

Reply via email to