Proposals and Research

This section collects official forward-looking CQLite proposals. It is not a raw research notebook. Pages here summarize the direction the project is prepared to stand behind, then link to the deeper source material for readers who want the full trail.

Status legend

Status	Meaning
Accepted direction	The architectural direction is the project plan, though implementation may still be staged.
Proposed spike	The approach is specific enough to evaluate or prototype, but not yet committed as product behavior.
Deferred	The idea remains documented, but is not part of the first implementation pass.
Rejected for now	The option has enough downside that it should not steer near-term work.

Research topics

Storage Engine Direction Why CQLite should sit beside Cassandra as an analytical plane over SSTables, instead of replacing Cassandra's internal storage engine.

Memtable Freshness How the project should close the gap between flushed SSTables and Cassandra's live unflushed tail.

Lakehouse Materialization How delta envelopes become queryable Iceberg tables without forcing every consumer to own merge logic.

Cassandra Seam Inventory The seam map behind the adjacent-OLAP decision: what Cassandra lets us plug into, and what remains hardwired.

Read Path & Query Providers How the Flight, Trino, DataFusion, and partition-read enhancement issues fit into a single query-provider direction.

Compaction & Maintenance How standalone compaction, explicit gc-before / now-sec inputs, and maintenance scheduling fit together.

How to read this section

Public first

Each page leads with the decision, trade-off, and current status before linking to raw notes.

Research-backed

Source reports stay linked so readers can inspect the evidence and the discarded paths.

Diagram-led

Diagrams explain architecture and data movement with hand-authored labels. Generated images are used only for non-critical conceptual visuals.