Submerge

Experimental database thing

This is a very early and experimental sketch of a distributed database. Justification and explanation will be forthcoming if it ever amounts to anything. Not serious yet.

Slogan:

"A 21st century 4GL from a parallel universe where the web didn't win"

Rough picture:

• Nested-relational data model
• Tiered local storage
  • Small BTree tables for hot data
  • Large LSM slabs for cold data
• High-consistency replication and commit protocol (Ocean Vista)
• Low-consistency replication and commit protocol (Atomic CRDTs)
• Typed query language on top (vectorized Ei calculus)
• Incremental evaluation on top (Dyn-FO + staging)
• Version control on top (nested branch-stage-update-commit)
• Automated QA, workflow triggering and CI framework on top
• Information-flow provenance/integrity/confidentiality
• Continuous incremental cloud backup and replica-provision
• End-to-end "full application" support:
  • Integrated simple auth and admin
  • Integrated simple multi-platform UI
  • Integrated simple billing / payment
• External-system interoperation:
  • Pub/sub between administrative domains
  • Format and protocol adaptors for web & legacy systems