Re: [rtl] IP useage from rtlinux, reply to Richard Teltz

[Paul Koning <pkoning@xedia.com>]

>>>>> "Thomas" == Thomas Wuensche <wuensche@ems-wuensche.com> writes:

 Thomas> Richard Teltz wrote:
 >> ARCNET, CANbus, etc are not supported by a terrifically wide
 >> number of vendors, and QNX is great if all your machines are QNX
 >> (kind of like MicroSoft).

 Thomas> What is your background information to state that CAN is not
 Thomas> supported by a wide number of vendors? This may hold true if
 Thomas> you compare CAN with the whole range of office applications,
 Thomas> if however you compare realtime related applications, I think
 Thomas> it is wrong.

 Thomas> However it seems to be an excellent idea to use CAN for the
 Thomas> time critical lower layers and Ethernet for the less critical
 Thomas> higher layers. Their basic transmission concepts are similar,
 Thomas> however CAN avoids the collision problems, which make
 Thomas> Ethernet non-deterministic.

I've already commented on the fact that calling Ethernet
non-deterministic is misleading.

Apart from that, I'm puzzled: this note says that "CAN avoids the
collision problems" while another note says that CAN uses addresses
(station IDs) to control who backs off in a collision.  So does CAN
have collisions or doesn't it?  And what about the unfairness of using 
addresses to prioritize media access?  (One of the key goals for any
LAN is fairness; DQDB had problems being widely accepted in part due
to its lack there.)

	paul
--- [rtl] ---
For more information on Real-Time Linux see:
http://www.rtlinux.org/