- make Record searches return multiple values
- create entity-reference adapter
- return type minimally
- if possible, query for label
- make sure that fields are the same as entity (juno table can't have multiple types)
- add Record title to Metadata vertex?
- all types are atom strings, but should be refs to types
- standardize schema/data model across sources
- standardized approach for making backwards-compatible schema changes
- parse produces a graph, not document
- use avro to serialize parse output
- schema is based on a config, not build into code; housed in a service
- can we take an Avro schema and produce instructions for
- how ElasticSearch indexes the fields?
- how to map fields to labels?
- other requirements?
- can we at least agree on the need to move beyond collider?
- metadta
- primary key
- primitive type
- field key, Json key, Bson key (not needed?)
- elastic field name and type (text, int, etc), and options (boost, analyzer,)
- label
- Keyword?
- DisplayUrl
- how to translate a field
- if a field can be geocoded
- if a record's date can be determined by a specific field
- schemaless fields (fields that don't make it into base ontology) will still need definitions for how to visualize it (see metadata above)
- e.g. Slovakia
Registered Seat
- e.g. Slovakia
what was the final suggestion by Farley?
how to handle Record versioning?
- can we be proactive on this
- forking/R&D
- transition plan?