John VanLoon commented on:
>[EMAIL PROTECTED]] wrote:
[ ... long posting in favor of a HAL/RTAI for RT-Linux ... ]
Gentlepersons;
My experience for many years has been developing real-time control
systems for large expensive (>US$1M) semiconductor process equipment.
This is a business environment where there are no acceptable excuses
for failures, especially software crashes.
The RTOSes I have used in the past have been commercial offerings
and in several cases paid the extra money to get the source
code because we needed it to keep out project out of the ditch.
>From that experience has come a list of software environemnts I
will never touch again and the reasons for that opinion.
I find RTL very attractive for all of the obvious reasons, but for
my comfort I need to see two things.
1. Certainity (as best as can be had) that Linux portion of the
OS code contains no suprises. In the Linux development model
the opportunity to mess up is non-trivial. This means to me
that RTL should not be chasing the latest offerings from Linus
and company, but rather focus on more deliberate changes.
Toward that end I would like to see as few releases as is
consistant with bug fixing. Let the improvements (new features)
come in larger chunks.
2. Now to the topic of this thread..., I too would like to see a
clean seperation between the Linux and RT-Linux code bodies.
If nothing else it would help make changes in the RT and non-RT
bodies of code independent, which can only help with the control
of the total code as a package.
Both of the above would have a positive influence on the content
(reliability) of the OS code. The underlying problem being that
Linux is too big to just read the code and know that it is OK.
It is these sorts of issues that must be addressed if RTL is to be
accepted in the industrial world where a bad decision can take your
employer into a losing Quarter. Such events are career ending for
the managers and VPs involved. The s/w engineers involved can
usually move on, but the wheels will be tagged with their history.
At the most recent Linux World in San Jose, VK asked the BOF
attendees if certification was an issue for RTL. My answer is
that certification can provide the necessary controlled source
evolution by accepting changes from the on-going Linux development
only after a satisfactory amount of experience has been gained.
IMHO, better source control of RT-Linux is essential if it is to be
taken seriously in the REAL real-time world.
I really Really REALLY would like to see RTL become something other
than a labratory curiosity. This statement is NOT an insult. It
is rather a statement of present reality.
Regards all,
Ray
PS: BTW, I haven't downloaded RTL and played with it myself, simply
because I lack the spare time to do it. I have read most
everything on the web and am quite impressed. I am ready to
use it for a lab project to gain the experience when next one
presents itself. I think a production project will need to
wait a while. Maybe next year things will have settled down
a bit.
--- [rtl] ---
To unsubscribe:
echo "unsubscribe rtl" | mail [EMAIL PROTECTED] OR
echo "unsubscribe rtl <Your_email>" | mail [EMAIL PROTECTED]
----
For more information on Real-Time Linux see:
http://www.rtlinux.org/~rtlinux/
