For large datasets we will likely need to provide a way to access and manipulate relational databases. Ideally, this capability would be implemented in parallel with (and not instead of) other database support mechanisms that support more flexible schemas (see Simple database support). This tool would initially be oriented towards generating web UIs for accessing and updating sql tables.
Issues include:
- Security. For databases that are created within the system, CORK/BRIDGE security restrictions should be enforced.
- Garbage collection. Unused databases (created within the system) should be removed when the corresponding bridge objects are deleted. The (forthcoming) garbage collection mechanisms used for SharedFiles should be sufficient for this.
Object types could include:
- Database objects
- External database objects would contain meta-data such as host, port, database name/path, user name, and password
- Internal database objects could be managed internally on the bridge server. The interactive UIs for these could allow management of database schema.
- Query objects. These would wrap sql SELECT statements, providing table views in the interactive client and table or list views on the web.
- Tables should optionally have edit links, with associated forms
- Form objects for inserting or editing
- Rows need unique id field for editing?
- Could support javascript-based field validation
- Need to be able to specify required vs. optional fields
- Need success/failure actions
Links:
|