ButlerWG Meeting 2017-09-05

Meeting began 05 Sep 2017 @ 11:30am; ended 12:50pm Project Time.

Attending

• Tim Jenness
• Jim Bosch
• Simon Krughoff
• Russell Owen
• Unknown User (pschella)
• Brian Van Klaveren

Next meeting: 07 Sep 2017 @ 10am.

New Use Cases

• We discussed some new use cases. Calibration Productions team need EFD access during batch processing. This was already part of DRP2 but has been explicitly called out as a calibration use case.
• DRP14 is to be modified to more explicitly describe how the DRP interacts with L1 data products. In particular, there is no need for DRP to see the updated WCS information, merely for the updated WCS to be available to the system working out whether specific CCDs are needed for a coadd.
• Tim Jenness noted that raw data downloaded by the end user will not be the same file as that stored in the data back bone but will have enhanced metadata content with a full set of headers rather than the minimalist set from L1, possibly with updated values, and possibly including updated WCS. It seems likely that this will have to be done as the raw raw and a separate metadata file as for composite datasets.

• Tim Jenness Write a use case involving raw data enhanced metadata.

• Simon Krughoff mentioned some new use cases:
  
  • A need for subsetting though a remote server in order to minimize local resource usages (see also DRP8e).
  • Interacting with VOSpace in a notebook and then having a VOSpace-aware butler on laptop.
  • A notebook running on some older data. How does the notebook trigger retrieval from tape?
• Brian Van Klaveren wondered if a DAX use case should be written using VO data access to arbitrary resources, leveraging ObsTAP and VO data models. Tim Jenness suggested that it should be included even if later we deprioritize it.
• Russell Owen was concerned that support for HSC and DECam for KPM verification and validation was not being explicitly mentioned. Jim Bosch agreed and updated wording.

Context Matrix

Jim Bosch was concerned that the rows are too specific. In particular "verification cluster" is not what was really meant and is a construction concept, and the multi-core rows might be better-named workstation and laptop. Simon Krughoff responded that he concurs that some naming can be improved, the important thing is to understand whether the gaps in the matrix are important.

Actors

We spent some time reorganizing the actor tab in the spreadsheet and grouping hierarchically. Also started adding software agents. Use Cases should be updated to use standard actors by next meeting.