2020-09-28: Development Meeting

Kacper, Tommy, Mike H, Mike L, Tim, Craig

Agenda

  • Updates
  • Discussion on provenance storage – [name=Tommy]

Updates

  • Kacper
  • Tommy
    • More recorded run docs, mostly finished this document
      • Would like to discuss after the Updates
      • Would also like to circle the wagons on it on Friday
  • Mike L
    • Closing on my new house this week :tada: :confetti_ball:
    • Just got deploy-dev set back up with virtual_resources branches
      • NOTE: Something in master (wt_home_dirs / girder_wholetale) is no longer compatible, this forced an upgrade
  • Mike H
  • Tim
    • Blazegraph can be configured to infer the closure of implied triples based on RDFS (e.g. rdfs:subClassOf) or (simple) OWL (e.g. owl:sameAs, owl:inverseOf) triples (Blazegraph iterates until fixed point is reached).
    • This is done (mostly) when the data is loaded into Blazegraph, rather than at query time (there is no full-blown, i.e. mega-slow, reasoner at query time), so triplestore will be big, but queries will be fast.
    • Nice because we can define our own nice, easy to query vocabulary and let others query it (with greater complexity) using PROV or ProvONE vocabularies.
    • Currently defining a “Comprehensive Provenance Record” vocabulary (prefix ‘cpr’) that declares subclass, synonym, and inverseof relationships to PROV and ProvONE. Provenance captured from ReproZip will be inserted as cpr triples.
    • Whole Tale (prefix ‘wt’) and SKOPE can derive any additional terms required from cpr. Queries phrased in terms of wt, cpr, provone, or prov vocabularies will then work.
    • In theory, exporting ProvONE triples (or even just vanilla PROV triples) will be very easy. They will already be in the triplestore and can be pulled with a simple CONSTRUCT query.
  • Craig
    • Deployed v0.9.1 with new infrastructure (TACC, Ubuntu, more storage, Docker volumes)
    • PR review
    • Quarterly reports
    • This week:
      • Need to get back to C2metadata
      • XSEDE XRAC