[Xenomai] [RFC] RTnet, Analogy and the elephant in the room

Philippe Gerum rpm at xenomai.org
Tue Nov 21 18:11:52 CET 2017

As the GIT commit history shows, there has been no sustained effort in
maintaining several of the Xenomai I/O frameworks for several months
(e.g. RTnet, analogy), even years in some cases, despite obvious bugs
are still haunting the code base according to the mailing list. The
situation has reached a point where I see no alternative to dropping
them if the situation does not improve, because there is absolutely no
point in this project shipping bit rot software that won't ever be fixed.

Unfortunately, this is only the tip of the iceberg. Let's face it, since
Gilles passed away last year, I have not been scaling to the Xenomai
maintenance, development and documentation tasks, with the requirement
of running my business in parallel.

The solution to this serious problem is fitting the project to the
available resources by narrowing its goal, or conversely, by growing the
pool of contributors.

In addition, we can rework the most tricky part of the implementation to
make it simpler to maintain, better documented, drastically lowering the
barrier on entry for new contributors, which is what I've been working
on for a year with the 4th generation of the interrupt pipeline.
Progress on this front has been significant already, but once again
limited by the time I have been able to devote to this development so far.

For the past 16 years, this project has lived on various types of
contributions from only a few committed people and companies. At this
chance, let's mention that people who have been deploying Xenomai in
industrial applications owe a lot to Wolfgang Denk from Denx
Engineering, Siemens's Jan Kiszka and Jorge Ramirez, who have supported
the project in crucial ways directly or indirectly over the years, and
still do.

Xenomai as a dual kernel technology showcase has been quietly delivering
on the promise of real-time Linux for more than a decade now, with
marketing tools limited to showing decent code quality, good and
reliable performance figures. As a result, to my current knowledge,
Xenomai is present in a broad range of applications and systems:
magnetic resonance scanners, 2D/3D printers, navigation & positioning,
communication equipment, autonomous vehicles, control & automation
systems in various plants.

On the sad side of the story, this project has virtually become a
one-man show the day Gilles - my long-time friend and hacking soul mate
- left us. That show is too big for me to run it alone, which entails

- the interrupt pipeline for 7 CPU architectures
- the Cobalt co-kernel
- 4 APIs, plus the "copperplate" mediating layer
- several real-time I/O frameworks, including CAN, RTnet, Analogy, SPI, GPIO
- the documentation (which is currently unfriendly to newcomers, and
stalled two years ago or so)
- the website
- the testing and release processes

So, let's talk about the elephant in the room: the current situation of
the Xenomai project is not viable in the long run. I can only encourage
people who feel concerned about it to discuss openly the practical steps
to best address this challenge.



More information about the Xenomai mailing list