Experimental Physics and
Industrial Control System

Ben Franksen <[email protected]> · Sat, 9 Oct 2010 01:24:46 +0200

On Samstag, 9. Oktober 2010, Patrick Thomas wrote:
> The channel access reference manual states that:
>
> There is unfortunately no field in the protocol allowing the server
> to indicate that an intervening subscription update was discarded.
>
> Is there a way to ensure that an intervening subscription update will
> not be discarded? Would you have to ensure that each callback
> executes in a certain amount of time, ie, faster that the channel
> update interval?

Relying on timing constraints for correctness of code is strongly 
discouraged. There are just too many factors which you cannot reliably 
control that can influence the timing, most importantly network 
latencies, but also delays in the kernel. For instance, the version of 
vxWorks we are (still) using (5.4.3) has the nice property of delaying 
all incoming traffic very 5 seconds for a noticeable period (I cannot 
remember how long exactly) for whatever reason, probably checking if 
the cable is still connected or something similar stupid.

The only reliable way to ensure that events don't get lost is to use a 
handshake protocol (or something more optimized, like TCP). One cheap 
trick to at least detect (reliably) if and when events get lost is to 
make sure (on the IOC) that your events alternate with another event 
from a counter.

Cheers
Ben
-- 
"Never confuse what is natural with what is habitual."
                                                 -- Mahatma Gandhi

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System