Here are my notes from Tuesday's discussion. Please correct/augment as necessary.
We discussed and decided on a number of issues stemming from Frederik's first round of feedback on the API (https://github.com/RDACollectionsWG/apidocs/issues/1)
PID Assignment: an implementation of the API should:
- include a default provider for assigning PIDs which implements the PIT API
- allow for a client to supply an endpoint for an alternate PIT API provider as a parameter to the CREATE request for a Collection
- declare in a capabilities call to clients what the default provider is and what providers can or cannot be used (i.e. a whitelist/blacklist)
- current operations to get the list of supported access types and model types should be assumed under aforementioned capabilities operation
- additional capabilities to declare include the max limit on # of collection items which can be operated on in a single request (i.e. CREATE, UPDATE, DELETE)
- POSTs of new Collections should return a single item, not a Cursor
- Requests which return a set should return a cursor
Transactional and error handling questions on adding multiple items at once to a collection
- we will restrict scope to support only synchronous requests
- POSTS which operate on multiple items where not all succeeded should return a FAILED status and provide some way for the client to know which failed and which succeede
Responsibility for dealing with recursiveness (collections of collections) is pushed to the client ? [ my notes on this point are fuzzy ]
Other items: Tobias will apply for a session at P8
Please chime in with whatever I missed.