Context and Framing¶

The discussion centers on rethinking scientific publishing infrastructure—moving beyond PDFs and disconnected supplementary materials toward modular, interoperable, and machine-readable research objects. Participants refer to this initiative as Bedrock, a metaphor for the underlying content layer of science that supports higher-level representations (“soil” as metadata, “flowers” as the presentation layer).

The group’s goal is to define or prototype a schema or standard format that enables linking, reuse, and interoperability of granular scientific content—figures, data, methods, and even paragraphs—across tools and platforms.

Problems in the Current System¶

Disconnected Elements and Reproducibility Gaps¶

• Researchers cannot easily rebuild or replicate another scientist’s work because key contextual elements (e.g., reagents, SKUs, metadata, data availability) are not structured or linkable.
• Supplementary files are idiosyncratic, unstructured, and inaccessible.
• Peer review focuses on narrative, not replication (“peer replication” was emphasized as the real goal).

Authoring Pain Points¶

• Scientists are not incentivized to produce structured metadata; it isn’t useful to them during creation.
• Without author-friendly tools, metadata curation will not occur.
• Late-stage publication (a year after experimentation) leads to data loss and forgotten details.

Cultural and Structural Barriers¶

• Researchers fear being scooped if they share granular results.
• Incentives reward polished narratives over incremental contributions.
• Big-science examples (e.g., LIGO) show structured collaboration and replication mechanisms that small-scale labs lack.

Desirable Properties of the New Approach¶

Modular and Granular Content¶

• The minimum unit of contribution could be a single experiment or even an observation (“the minimal currency of science”).
• Each unit should be typed and linkable—a figure panel, paragraph, dataset, or method.
• Enables peer replication and building cumulative knowledge graphs.

Author-Centric Design¶

• Metadata entry should be embedded early in the research workflow (lab notebooks, weekly logs).
• Tools should give immediate value to authors (e.g., better organization, feedback, or linting-style suggestions).
• Shorter publication cycles (iterative curation) reduce cognitive load and error propagation.

Linking and Web-Native Infrastructure¶

• The web’s power lies in linking, not redisplaying.
• Embrace messiness and heterogeneity—interoperability over standardization.
• Support bi-directional or resilient links between research objects while accepting web fragility.

Layered Metaphor¶

• Bedrock: raw content and data.
• Soil: metadata and schemas enabling discovery.
• Flowers: presentation and visualization layers that remix and display content differently.

Technical Architecture and Standards¶

JSON, JSON-LD, and JSON Schema¶

• Core consensus: use JSON as the base data format.
• JSON Schema defines structure and typing (paragraphs, figures, links).
• JSON-LD adds linked-data semantics—shared vocabulary alignment across schemas (via @context).
• Schema.org terms and JSON-LD mappings enable translation between domain vocabularies.

Relationship to Existing Standards¶

• Pandoc AST provides a precedent but is “hideous” and not ideal for machine processing.
• JATS is acknowledged as important but overly rigid and non-interoperable; the group seeks a successor schema more like JSON/HDF/Zarr—modular, extensible, and linkable.
• Compatibility with MyST, Stencila, Quarto, Curvenote, and other document models is critical; they are viewed as “interfaces” over a shared data model.

Linking Mechanisms¶

• Every element (paragraph, section, figure) should have a persistent ID or resolvable path.
• External references (e.g., datasets, code) would use typed links validated against schemas.
• Agreed-upon “carve-outs” for cross-references in ASTs.

Incremental Adoption¶

• Consensus: don’t wait for universal agreement.
• Start with two or three collaborating projects, publish version 0.1 of a shared schema, and test interoperability.
• Translate existing documents (e.g., JATS → JSON) to bootstrap adoption.

Licensing and Attribution Layer¶

Machine-Readable Licensing¶

• Licenses (e.g., CC-BY, CC-BY-NC, ND) can be expressed and validated mechanically within schemas.
• Linking is permissible even between incompatible licenses (not a derivative work), but reuse and remixing require compatibility checks.
• Future tooling could include a license-compatibility matrix or automated validator.

Attribution Requirements¶

• Schema can embed rules and tooling to enable “if you link to this, include this attribution and license”.
• Encourages early sharing by providing attribution safety—time-stamping and provenance metadata (a digital “flag-planting”).

Creative Commons Perspective¶

• CC representatives emphasized user awareness of license implications and limits of automated enforcement.
• Mechanized verification is feasible and desirable but must complement, not replace, human interpretation.

Broader Philosophical Insights¶

• Continuous Science: Move from static, monolithic papers to continuous, linkable research streams.
• Trust Circles: Allow sharing and collaboration at intermediate stages without public release.
• Open by Design: Machine readability and openness should be defaults; opacity should require justification.
• Messy Web as Feature: Embrace diversity of tools and formats; prioritize translation and interoperability over control.
• Schema as Conversation: Rather than one fixed format, define shared translation points between evolving community schemas.

Next Steps and Outcomes¶

• Draft a proof-of-concept schema (version 0.0.1) defining minimal elements:

  • Paragraphs, sections, figures, datasets, and cross-references.
  • JSON Schema typing plus optional JSON-LD context.

• Prototype translation between MyST AST, Pandoc AST, and Stencila Encoda.

• Define a mechanical validation layer for licensing compatibility.

• Publish demonstration documents across platforms (Curvenote, Quarto Pub, eLife, etc.) to test interoperability.

• Continue collaboration toward a “Bedrock” format that underlies composable, modular, and interoperable scientific publishing.

Key Takeaways¶

1. The problem is structural: PDFs and JATS lock science in rigid, narrative-centric silos.
2. The opportunity is architectural: A web-native, JSON-based schema can unlock modular reuse.
3. The transition must be incremental: Begin with partial interoperability among active communities (MyST, Stencila, Curvenote, JATS, etc.).
4. Success is cultural as much as technical: Scientists must see value in the format immediately.
5. Licensing and attribution are enablers, not afterthoughts: Machine-readable rights are core to open, reproducible science.

The Bedrock conversation charts a pragmatic path from today’s disconnected scholarly formats toward a composable, interoperable, and link-rich ecosystem for scientific knowledge—grounded in JSON schemas, aligned through linked data, validated through licensing logic, and grown organically from community-driven collaboration.