Content Hosting (CH) is currently implemented via a two-tier set of objects: one to store content and properties in a datbase, the other to store content in files. If content is saved in files and properties are kept in the database, this presents a serious problem with resource integrity. For example, if a file is deleted in Sakai, and the user asks their IT people to restore it, the metadata is lost, potentially corrupting the Content Hosting system. Also, the implementation uses an internally defined method of naming files and directories that makes restoring files nearly impossible.
Given a desire to support restoring lost files and external file system management (via AFS, etc.), I am exploring a new implementation of Content Hosting based solely on a standard file system. Naturally, there are trade-offse with this approach. It may not be as fast as the existing implementation, and some way needs to be found to associate base files with their metadata files. Coupled with this is a desire (expressed by some) to implement some form of JSR-170.
Planned Approach
- Map directory names and file names directly to an underlying (specified) file system.
- Save properties in a file called file-name.properties, where file name corresponds to the base file name.
- Save attachement references in a file called file-name.attachments.
Issues
- Properties can only store string-based properties.
Project Status
- Implementation of entity interface on Resource has been started.
You can track progress via my CH Project Notes.
Things to Do
- Move the work to contrib.
Joining
Send me email if you are interested in participating in this project.