In no particular order:
- define an easy to use data access and mesaging layer √
- define a model structure that avoids the duplication we have now √
- land on better administrative and public api's (with openapi)
- generate/run public discovery website from the same codebase that runs the backoffice
- better search, own query language
- (re)index pipeline √
- runtime index switching √
- use opensearch versioning √
- avoid deep paging √
- split people and users
- api user permissions (jwt)
- store affiliation duration
- configure http caching correctly
- representations/citations table √
- integrate oai √
- integrate citeproc
- maintain fewer support libraries
- strongly typed translations
- key rotation