Data projects

[tdudgeon]

I think we need a new concept in the data tier - a project.
This will allow better collaboration between users.

• A dataset belongs to a project (potentially to multiple projects).
• A project can be personal or shared.
• A personal project can be changed to being shared but not vice-versa.
• Only an owner of a project can manage datasets in that project.
• A shared project can have multiple owners. An owner can add another user as an owner.
• All users can view and use data in a shared project (later we may add access control).
• Datasets can be listed, filtered based on project.
• Each user starts with a default personal project that has the same name as their username.
• The API should list the available projects.

A consequence of this is that datasets are not really 'owned' by an individual user, they are 'owned' by the project.
Instead each dataset is 'created by' an individual user and as datasets are immutable (other than deleting them and updating metadata) then this will never change.
In terms of filtering, filtering by project is typically more useful than filtering by user.

[alanbchristie]

A shared project can have multiple owners

Is there a concept of something between personal and public? i.e. a closed or restricted project? i.e. one that's not public but may have a number of users or owners that are allowed to use its data?.

[alanbchristie]

Datasets can be listed, filtered based on project.

What impact does this have on the existing /data GET endpoint and the original MVP, which filters based on User and Label?

Is the User still necessary when it comes to searching for data?

Also what do Labels apply to? Can they be applied to all objects (i.e. User, Data and Project)? Or do they now just apply to projects?

Is this an extension to the MVP that we can deal with later or are projects part of the MVP (#20)?

[alanbchristie]

Each user starts with a default personal project that has the same name as their username.

This is too complicated to implement in the backend autonomously. A user isn't known until they use an API method and to create a project automatically from the first invocation of any method will probably just result in messy code. If this requirement is needed then it needs to be a responsibility of the application layer, not the API. But, do we really need this? If a user needs a project then they can create one, with a name of their own choosing and we can avoid this special case.