Tackling the Write-to-Read Web of Data with Trustflows

Ben De Meester, Julián Rojas, Femke Ongenae, Pieter Colpaert, Ruben Verborgh

IDLab, Department of Electronics and Information Systems, Ghent University - imec, Ghent, Belgium

⚠️

The read-write symmetry in current decentralized ecosystems for Linked Data hinders interoperability

  • What you write is not necessarily what you want/need to read.
    • The data models needed to read data diverge from the data model used to originally write the data.
  • Data models change and your app needs to follow.
    • CRUD symmetry forces apps to individually interpret evolving data models into their own context.
  • Wait! Who is in charge of these data?.
    • When multiple actors can write data, the implicit trust context of a single authority is no longer valid.
✅

The Trustflows approach: CQRS + Event Sourcing for explicit Trust contexts

Architecture

Trustflows operational model

  • CQRS (Command Query Responsability Segregation) to decouple writing from reading.
    • More flexibility at the cost of increased complexity.
  • Common semantic mappings to handle data model change.
    • Catering for evolving and alternative data models .
  • Event Sourcing for explicit trust contexts
    • Provenance (e.g., PROV-O) is made explicit for every data write.