>From: "Dresner, Norman A." <[EMAIL PROTECTED]>
>Subject: RE: [rtl] Networking With RT-Linux
>Date: Tue, 23 Nov 1999 11:56:54 -0500
>
> There are a large class of networks that are designed to have
>"errors", networks like Ethernet that work by collision-detection.
>
> But there are a larger class of networks that have "no errors"
>during normal operation, for example networks that work on master-slave
>protocols and token-ring networks. {Also MIL-1553x MUXBUS networks which
>I'm heavily involved with}
That's definitely NOT true. Any communication protocol carries an
intrinsic non-zero error rate (as was well said before). Token-ring
networks can be used where to allow a deterministic behavior, but that
has nothing to do with the error rate.
The collision detection in CSMA/CD (ethernet for one) networks just
makes it non-deterministic, but the error rate can be smaller (or greater)
than a token-ring. That's a complete different issue.
> The consideration of whether master-slave networks are
>"deterministic" or not depends on your point of view, i.e. whether you're
>the master or the slave.
Again, that's not true. Both, master and slave can be guaranteed
to be deterministic (or guaranteed to "talk" in deterministic intervals).
This is NOT limited to whether the node is assigned to be master or slave.
> GPIB is also master-slave, but not deterministic by design because
>the timing restrictions on the "talker" or listener" (the slaves) is
>significantly looser (millisecond rather than microseconds) but still there
>is a computable maximum time interval (ignoring the complication that a
>talker-slave can transmit as much data as it wants on the bus).
I think that there is a big misconception about what is real time
and what is a fast response time (processing time, etc). Real time is
related to determinism, rather than to response time. A system may
present a response time of 1sec and still be real-time (deterministic)
One may design a "real-time" system to control a clock. It requires
a response time of 1 sec. But it doesn't matter if it is fast or not.
Only that it is guaranteed to be "available" every second.
> Token-ring networks are inherently non-deterministic (within sloppy
>limits, I suppose, depending on the specific network protocols governing how
>long messages can be and how many messages a station can transmit before
>passing the token to the next station),
Again, I may be wrong (have forgotten my network protocol class), but
as far as I remember, token-ring networks ARE deterministic. Every message
can be guaranteed (probability one) to be delivered within a fixed number
of cycles (passing of the token).
> But there is assumed to be (under normal operation) no errors during
>network operation because there is (again under normal operation) no
>possibility for collision.
Collision is not the sole source of errors in a communication protocol.
If you ever used a ppp connection (point-to-point implies only two nodes
with two independent wires, one for tx and one for rx, and therefore not
chance for collision) instead of ethernet (TCP/PPP with or without modem
- dialup server in ISPs) you could easily confirm that.
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/
