15 hours ago
- Files in personal computing are owned by users, not apps, and outlive the software that created them.
- The social filesystem extends this concept to social apps, where user data (posts, likes, follows) are stored as files in a personal repository.
- Records are JSON files organized into collections, with lexicons defining the schema for each type of record.
- Identity is handled via DIDs (decentralized identifiers), which allow permanent links that survive changes in hosting or handles.
- Apps are reactive to file changes; derived data like like counts are computed from records across users' repositories.
- Repositories are self-certifying via hash trees, and relays enable streaming of changes for app-specific caches.
- Real-world examples include Bluesky, pdsls, Sidetrail, and teal.fm, demonstrating the feasibility of the approach.