I am not sure I got this thread from the beginning, but...
>Date: Mon, 22 Nov 1999 08:56:54 -0700
>From: [EMAIL PROTECTED]
>Subject: Re: [rtl] Networking With RT-Linux
>
>In practice this is absolutely correct. There are 1000 microseconds in
>a millisecond, the more of them that are allocated to non critical
>tasks, the less are available for critical tasks. If you make everything
>"realtime", then nothing is realtime.
Definitely correct! If a rt task takes too long to run, and mainly,
if takes non deterministic time to run, all other rt tasks are
jeopardized.
>Linux TCP/IP is a very well designed and enormously complex system. It would
>be hopeless to try to recreate it at the RT level even if we ignore the
>fact that TCP, which most people want to use, is fundamentally not
>realtime.
Again, non determinism at RT level contradicts everything else in RT.
TCP is notoriously non deterministic (domain resolution, loss of messa-
ges, CSMA-CD protocol itself). I for one don't see any sense in having
it "real-time".
>Many RTOS's include much more than RTLinux and they are slower, less
>predictatable, and more difficult to maintain.
>
>The RTLinux design paradigm puts time sliced components on the Linux side
>which is optimized for time-slicing. The RTLinux side is optimized
>for low latency. There is a big difference.
Yeap, but maintenance is always a big issue in RT systems design.
I've seen many people claiming to do RT applications and having huge
processes running at the interruption handling level. It just don't
make sense.
Guilherme
-------------< G. N. DeSouza >-------------
---------< [EMAIL PROTECTED]>---------
--< http://rvl1.ecn.purdue.edu/~gnelson >--
--- [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/
