Experimental Physics and
Industrial Control System

Claude Saunders <[email protected]> · Thu, 20 Oct 2005 09:19:29 -0500

I'm of the opinion that a device layer needs to be a separate namespace

entirely, not just aliasing within the same (CA) namespace (for the same

reasons Ralph raised). That's not to say it still can't be a part of CA

server, or a separate gateway. Aliasing within the same namespace is

mainly useful for protecting applications from changing underlying

implementations (as in the set of linked records to provide a single

value). We have such aliasing now with the gateway.

Secondly, I think one of the big benefits of a "device layer" is the

ability to bring to bear other resources such as a relational database.

So an atomic write to a [particular device].setpoint would actually wind

up kicking off a write to data access and a relational database (for

whatever reason).

Of course, what I'm suggesting is more complicated .... :)

   - Claude

Ralph Lange wrote:

Hmm...

You're mixing granularity levels: Nowadays every PV name that doesn't

contain a dot is regarded as a record name.

So, when you're aliasing [particular_device].Setpoint to

[particular_record].VAL any generic or display tool will soon start

asking for things like [particular_device].Setpoint.EGU .RTYP .DRVH

.DRVL .STAT .SEVR and so on and so on.

Do you want to create explicit alias definitions for all (pseudo)

fields of that (pseudo) device record? You will end up with zillions

of alias definitions on a single IOC.

If you try to escape by aliasing only on the same (record) level, you

don't gain as much. Simply having other names for records just creates

ambiguity - which of the many names for the same record do you use for

archiving? For snapshots? For the alarms? How do you compare snapshots

that contain the same record under different names?

Guess this needs more thought - maybe too simple?!

Ralph

Marty Kraimer wrote:

If I read this correctly it is just asking for the CA server to

support aliases.

Marty

[email protected] wrote:

Hmmm, a clever and simple idea.

  Ned

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System