Meeting began    @ 11:05; ended 12:45 Project Time

Attending

Next meeting  @ 10am Project

Requirements

We went through the remaining requirements with comments attached.

  • The "Registry of Data Repositories" requirements was accepted. Decided against explicit requirement for registries of registries or merging of registries.
  • We added use cases to "Data Output References"
  • "Provenance Graph" was removed as it is covered by REQ8a.
  • The "Provenance Tracing" requirement was updated by Michelle Gower to reflect that the requirement is on the data back bone provenance system and not a requirement that all data repositories have to understand how to trace from root provenance to products.
  • "Provenance Raw Data". This was a long discussion on whether this is really an "export" requirement on DAX to be able to add provenance to a file on retrieval. We did agree that writing of provenance could be treated as a composite dataset and that we might need a general requirement for exporting a composite made of many files into a single file. It was agreed that we are talking about the summary provenance and not the full provenance. Unknown User (pschella) and Michelle Gowerdid wonder how WCS provenance fits into this. Tim Jenness introduced the concept of primary and secondary provenance but we did not pursue this. Jim Bosch pointed out that we need guidance from DPDD as to how provenance should be handled from the user perspective: should provenance summary be in the downloaded files at all? Should we require that any provenance lookup must be done by querying the DAC?
  • REQ8a was accepted.
  • "Round trip time for DIASource and DIAObjects" was discussed. Is this a performance requirement on L1DB rather than on the butler side? Is it a DAX requirement? Kian-Tat Lim might have an opinion. Russell Owen reported that Kian-Tat Lim is expecting the butler to mediate the issuing of alerts and L1DB DIAObject puts. Russell Owen reminded us that DIAObject fetch can not be done in next visit pre-fetch (as for templates) because then you would miss DIAObjects from the current visit. Tim Jenness wondered if you could pre-fetch DIAObject information from the previous nights and then request only those new ones from the night. This was not discussed further.
  • REQ1999 needs input from Kian-Tat Lim.
  • Jim Bosch write a use case for dataset garbage collection. 
  • We discussed Simon Krughoff's LDF requirements. They generated some discussion as to the limits of the Data Back Bone and how much of the DBB infrastructure could be used for CI data products and provenance tracking. We decided to talk about this later as it's not directly impacting Butler design but does need to be resolved. Jim Bosch wondered if this would be covered by the ConOps document.
  • REQ542 was created by merging two requirements that referenced the base facility EFD.

Priorities

We discussed prioritization of the requirements. We tried to use the numbering of the scheme used by LSE-61 but with different definitions.

  • 1a: Core feature.
  • 1b: Needed by end of 2018
  • 2: Needed by operations start
  • 3: Nice to have

Everyone was tasked with adding priorities to the spreadsheet, in particular ensuring that the requirements derived from their use cases were covered. We also agreed that whilst someone can raise the priority of a requirement, you should not lower the priority that someone else made. We will discuss at the next meeting.

Design

Jim Bosch requested that for next meeting people read the Overview section of DMTN-056. He also stated that he would aim to have a first full draft by  afternoon and he would encourage people to read that before the next meeting if possible.

  • No labels