Why a portable ontology is the only way to escape vendor lock-in

Most data migrations hard-code the target dialect into the modelling layer. SCD2 is implemented in Snowflake SQL. PII masking is a Databricks UDF. The Gold mart's grain assumes BigQuery clustering. Switch targets and you're rewriting from the bottom up.

The ontology as a contract

LucidOntos models entities, relationships, history strategies, and PII tags as a target-neutral graph. The ontology never mentions Snowflake, Databricks, or any specific dialect. It's a contract between the source profile and the codegen layer.

Codegen is the swappable part

When the target changes, codegen re-runs. The ontology is unchanged. The mapping rules are unchanged. The pipeline shape is unchanged. Only the dialect-specific code is regenerated.

Reversibility is a property of the system

Because every artifact downstream of the ontology is generated, every artifact is reversible. Change the SCD2 strategy, regenerate. Add a new entity, regenerate. Switch a mart from Snowflake to Databricks, regenerate.