Experimental Physics and
Industrial Control System

"Kasemir, Kay via Core-talk" <[email protected]> · Mon, 14 Jan 2019 16:17:34 +0000

Hi:

Marty, Michael, thanks for the updates.

So https://github.com/epics-base/pva2pva/issues/11 already confirms that QSRV always serves the complete structure, and it'll be fixed when ".. changing QSRV to use SharedPV".

Good that you thought about the case where clients request some optional fields that may not exist.

> One of the selling points of pvData was the idea that extra fields could be
> added without breaking clients.  However, this break down if any particular
> layer is allowed to assume the it will see only those fields it expects.

Of course each layer needs to be somewhat tolerant.
Check if there is a "timeStamp", then get the "secondsPastEpoch" and "nanoseconds".
These two must exist, otherwise it's not a normative timestamp.
But if a server decides to add another field to the timeStamp, that needs to be tolerated by all clients.

Still, if I specifically request a field, it's problematic to send the complete structure.
The SNS neutron data clients request specific fields to reduce network load.
Sending the complete structure to all clients would overload the network.

For records and QSRV I don't see that as a practical issue because even though you get the complete structure, only changes require network bandwidth.
The only problematic scenario would be a large waveform record where some client solely requests the timestamp.
Always adding the complete waveform would increase the network traffic compared to just the requested timestamp, but I think that's a silly example.
In any case, sounds like you plan to update QSRV eventually, so then it's all perfect.

Thanks,
Kay 

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System