2019-08-19: Development Meeting

Kacper, Elias, Joe, Mike H, Tommy, Craig, Mike L

Agenda

  • Updates
  • Discussion
    • Tokens vs providers (comments here) – [name=Kacper]
    • Add girder SHA sum to manifest? Mentioned here ([name=Tommy])
      • Super easy
      • Any benefit though?
      • No
    • Discussion of whether bag artifacts (bag-info, checksum files, etc) should be added into the manifest aggregates section? ([name=Tommy])
      • Right now we add the top level README.md, run-local.sh, and fetch.txt but not any other files that we inject
      • Seems fine to me, but I just want to check for concurrence
    • Any prov stuff in particular that I should be working on right now? There’s a DataONE AHM and I don’t think much progress can be made on bag exporting this week ([name=Tommy])
    • Dashboard PR
      • [name=Craig] to review
    • Tale @ Scale discussion
      • To what extent should we be deploying containers in the spirit of reproducibility.
      • Can/should we be employing other methods rather than containerization

Updates

  • Kacper
    • PR reviews (wholetale#345, #342, #343; )
    • Meetings:
      • Tales @ Scale
      • Reproducibility and Prov:
        • Reprozip
        • Sciunit
      • Internal discussion about feasibility of Docker Images for archival purposes
  • Craig
    • A little time on Kubernetes migration
    • More time on meetings
  • Tommy
    • Aligned bag and zip manifest
    • PR for publishing bag artifacts to DataONE (run-local)
    • Discussing DataONE’s bag export (it looks like we’ll version it)
      • 90% sure we’ve decided on prov:atLocation for encoding file path
  • Elias
    • Reworking Agave provider PR (needs rebasing)
    • The HTTP Bearer Token PR is complete
    • Finished Stampede 2 tutorial and ran my own application
    • Starting work on learning Tapis/Agave to create an app using the API
  • Mike L
    • Reviewed Tommy’s PR - looks good :+1:
    • Angular refactoring: file browser restyled into a Table
      • Troubles encountered while wiring up the file browser
        • tale.dataSet does not contain file size - not performant to look up per folder/file
        • Too many models - collection vs folder vs file vs item vs dataset vs workspace etc
        • Endpoint confusion - it is difficult to write a reusable component when different contexts use entirely different endpoints and model formats
        • Each model is slightly different - difficult to write a generic FileElement model covering all cases
      • It would be worth having a quick chat about sane patterns for accessing Girder’s folder hierarchy
        • [name=Kacper] mentioned that Girder offers some helper methods for accessing folders/files that we currently don’t utilize
        • [name=Tommy] suggested taking a look his prior file picker PR
  • Mike H