Re: More flexible record scanning

[Andrew Johnson <anj@aps.anl.gov>]

Benjamin Franksen wrote:
> On Tuesday 28 June 2005 16:42, Andrew Johnson wrote:
>
> > What does a "Backward" link really mean, how do you implement it, 
>
> No, I meant a normal input link! You can make it CPP or CP, of course, 
> like any input link. Very simple.

"Backward link" is just my name for your input link; I don't want any 
common feature like this to only work with CA links (there should be 
nothing special about CA links in V4, they'll be just another kind of 
link support).  Your "very simple" implementation idea is guaranteed to 
only work with CA links that are CP/CPP, and I don't like that at all. 
To support it for DB links, we'd have to add all the stuff I talked 
about since we don't yet have change event notification on input DB links.

I really don't want us to add "Backward links" to EPICS; we should fix 
the event scan as the solution to your original issue.

That does bring up the point that there is a significant difference in a 
CA link that is CP/CPP than most other kinds of link: the "process me on 
change event" behaviour is unique for an input link, and we should think 
about whether we could have a standard way to support that for other 
link types.  There is something conceptually very similar about it to 
I/O interrupt scanning from an input device support -- could we create a 
single mechanism that implements both?  Maybe remove I/O Interrupt from 
the list of scan types even...

- Andrew
--
Podiabombastic: The tendency to shoot oneself in the foot.