Research and Use Standards

Rule

Before designing file layouts, conventions, integrations, or schemas, find the existing standard. Cite sources in the design doc. Recommend the standard unless you have a project-specific reason to deviate — and if you deviate, name the reason.

Why

Standards encode hard-won decisions that the field has already made and stress-tested. Reinventing them is expensive (the work itself), wasteful (every adopter has to learn the bespoke version), and risky (the standard usually handles edge cases the bespoke version misses). DPF’s strategy is to compose standards: IT4IT for value streams, CSDM for service models, TOGAF for enterprise architecture, MCP for agent tools, DCO for commit attribution. The platform’s leverage comes from being the substrate that wires those standards together, not from inventing parallel tracks.

Applies To

In-platform coworkers designing new features, external coding agents writing specs, and humans setting platform direction. Symmetric. Applies to data models, APIs, file formats, agent topologies, governance models, and integration patterns. Does NOT apply when the standards literature does not yet exist for the problem (early-stage research) — but the obligation then becomes to publish the bespoke design as a candidate standard, not to keep it bespoke.

How To Apply

For every new design surface, do the literature check before sketching code: find 2-3 OSS leaders and 2-3 commercial products, read their actual data models, document what they got right and where they got it wrong, name the patterns DPF adopts and the patterns it rejects. The “Research & Benchmarking” section is mandatory in every feature spec for exactly this reason. When deviating, the spec’s “Patterns rejected” section names the deviation explicitly.

Decision Dimensions

Examples

Sources

(Rendered from the sources: frontmatter by WikiSourceCitations — do not duplicate citation prose here.)