As soon as I saw Chris Albertson reply to Greg Bernard "Interesting GUI"
I knew what will happen because I posted somewhat similar message about
a year ago. Numerous attacks followed right away. The only surprise was
a size of the response this time so I decided to split from that thread.
Attacks from some individuals that are stuck in yesterday is very
disappointing. Others, including myself, that understand open source
software and related development are in support of what Chris is saying.
IMO it's extremely bad to attack those who come up with "it would be
nice to have this or that" in existing product. Saying, "it's open
source, go fork and write your own code" is plain STUPID! We are not all
programmers! [1]
[1] Before you (individual) attack me, let me tell you that I too worked
on computer systems in the early 1980s. Oscilloscope, soldering iron,
spare parts were my tools to get PDP-*, HP-*, S-100, Multibus, back in
functional order. Those were all industrial systems, some very critical
like X-ray machines, power plants, electric grids, etc.
Working on electronics side of computer systems kept me away from
software side so I never became a programmer. To learn at home I had to
smuggle my first ZX-81 and Amstrad (CP/M) behind the Iron Curtain.
As soon as the PCs became affordable and I made some money here in the
US I bought XT clone and started learning whatever was possible decade
before the Internet. One manager told me that my experiments with
introducing Linux to National Semiconductor in 1994 have no future.
Solaris was the king for engineers.
As technologies changed, my career veered into systems administration
supporting software developers mostly on Unix and Linux systems. Since
then I've seen my share of bad starts in coding projects. They are clear
indication of what code or style the first few programmers knew.
Lack of OS experience is the most dangerous one. "Running around with
Windows laptop" and using "porta"potty to connect to Linux is one
example I've seen too many times and it shows in final products all the
time.
---
I follow this LinuxCNC mailing list for years mostly as a systems
administrator. I would like to see and support it in industrial
environments. Unfortunately the demand is not there. When I talk to
potential users it becomes very clear that what Chris, Greg, and some
others were saying is true.
I paid for and supported open source software including buying Linux on
floppies and CDs. I sent donations to software developers and would not
mind doing so for LinuxCNC project (hardware and software) if it helped
me make money. If I had a chance to install it in an industrial
environment I would suggest the company contributed to
SW development. One of my managers paid hundreds for GNU CD package in
the mid 90s.
What LinuCNC is doing in its core was done on relatively simple systems
without GUI long time ago. I believe that it's time for architectural
change. Split GUI from RT section, and move away from the dependence on
terrible PC architecture to industrial SBCs make sense to me. Imagine,
we still have DOS functions in BIOS! And that is "emulated" in virtual
machines in the data centers these days!
Some SBCs are very expensive, others are cheap but poorly designed from
the electro-mechanical point of view. For example, RaspberryPi, BB, and
clones are good from programming side but not ready for industrial use.
Try to use scope on a middle board sandwiched between the RPi and
another PCB. What's most important is compatible interfaces developed by
different vendors.
Digital Corporation and HP encouraged that with open bus architectures
that handled multiple boards with well designed RT OS in PDP-* buses
over 40 years ago. Unibus, HP-IB, etc. I don't see that in LinuxCNC
world. It's great to see some people making $$$ from it but there is no
big industry or competition behind it.
One comment in that long thread mentioned whyfy as a totally unsuitable
way to communicate on machine shop floor. Time to read about net
technologies and related industry trends. Learn about 5G and how it
makes it possible to rearange CNC robots on the manufacturing floor
without rewiring the whole shop! Granted, we hobbyists and small machine
shops cannot afford this from the beginning but all things trickle down
eventually.
Another comment was negative towards python. Too bad. Python library is
huge which saves you from writing a lot of code. Python is used
extensively in Blender, ROS, graphical presentation of data, simulation,
systems administration, etc. Perl is dead. Most scripts I wrote in it
are gone with the companies.
I would rather see a resurrection of Multibus like architecture with
modern CPUs and peripherals for CNC use than running "special test" to
find out which freaking motherboard with parallel port on PCI card is
suitable for Linux RT kernel and LinuxCNC. Intel designed Multibus for
RT use and supported it for many years.
Most issues that were practically resolved in the 1980s come up again
and again; interrupts, real time, DIO, AIO, UI. Modern silicon solutions
can easily solve them in much smaller space as long as the starting
point is not a proprietary OS that made more damage to evolution of
technologies than earthquakes and typhoons.
All programmers should have good knowledge of computer history (1970s to
late 1990s) before they are allowed to write code for use in production
environments. Programs would be way more efficient, easier to debug or
troubleshoot, etc.
Is LinuxCNC going to stay as it is and end up as other old computer
technologies described in IEEE article "Inside the Hidden World of
Legacy IT Systems"
https://spectrum.ieee.org/computing/it/inside-hidden-world-legacy-it-systems
--
Rafael
_______________________________________________
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users
