It has been proposed that DPM needs to support settings
Erlang DPM's responsibilities are based upon CLIB-DPM's; handling readings and supporting redirection. Settings in CLIB may have shared some code with CLIB-DPM (database look-up, scaling routines, etc.) but the actual setting was done in the client's context. It has been suggested that the new DPM should handle settings.
- Clients would only have to track the latest DPM protocol to gain readings and settings. If DPM doesn't support settings, each client language needs to re-implement
- There has been talk about adding Channel Access to the bottom layer of DPM to support both ACNET and EPICs front-ends. If DPM handled settings, clients wouldn't have to know how to talk to multiple front-end architectures.
- DPM would forward settings to the setting service so clients wouldn't have to re-implement that code over and over.
What does it mean to run settings through a data pool? You can't merge requests.Settings won't get merged. We're just adding one more hop to the setting's path.The
SETS32header has an FTD field which can be used to schedule settings on an event, so the extra hop becomes irrelevant.
- DPM uses DRF2 strings to cleanly and succinctly describe a data acquisition request. DRF2 doesn't support the notion of settings. It would be nice if readings and settings had similar ways of describing the intent rather than having two unrelated APIs.
This topic needs more thorough discussion in a meeting.
#6 Updated by Richard Neswold about 1 year ago
- Status changed from New to Closed
Several of us had informal discussions that setting support shouldn't be part of DPM. In the past, the console DPM didn't making settings. Somehow we've attached this feature onto DPM. Moving settings into a separate service has the following benefits:
- DPM can focus on supporting all the ways we can read data
- Accelerator data from ACNET front-ends and IOCs
- Save/Restore data
- Data logger data
- FTP and snapshot data
- Setting service can focus on handing settings
- Protocol can be tailored to setting requirements (instead of shoe-horning it into the DPM protocol)
- Validating credentials of the request
- Writing successful settings to the setting database
- Handling knobbed settings
This codebase may actually run in the same VM as the DPM, but I'd recommend is not be part of the DPM package. It would be is own package and would connect to ACNET through its own handle.