Supported formats (version floor)

Issue #1249 codified this as a hard rule after roborev repeatedly raised “regression for older SSTable versions” findings on pre-na (Cassandra 3.x) formats — formats CQLite does not read at all. The churn came from dead version-gate modeling surface that looked like pre-na support. This page is the single authoritative statement of the support floor.

The rule

CQLite targets Cassandra 5.0. na+/nb BIG and oa/da BTI are in scope. Pre-na (ma–me, Cassandra 3.x) is out of scope and SHALL NOT be introduced, supported, or reviewed for correctness.

Format	In scope	Out of scope
BIG	na, nb, oa	ma–me (Cassandra 3.x) and anything < na
BTI	da	any version other than da

For reviewers (including roborev)

• Do not re-litigate pre-na correctness. A branch, gate, or test that touches ma–me is not a regression surface — pre-na reads are unsupported by design.
• Do not request changes that “fix” pre-na behaviour, add pre-na gates, or restore deleted pre-na modeling code. The floor is the contract.

How it is enforced

The floor lives in version parsing and is propagated by the reader:

• BigVersionGates::from_version returns Error::UnsupportedVersion { version, floor: "na" } for any BIG version < na. It never constructs gates for a below-floor version.
• BtiVersionGates::from_version returns Error::UnsupportedVersion { version, floor: "da" } for any BTI version other than da.
• SSTableReader::open propagates Error::UnsupportedVersion instead of degrading to the nb fallback. Only a structurally-unparseable descriptor (not a parsed-but-below-floor one) may use the fallback.

The rejection happens before any row data is read. Error::UnsupportedVersion maps to the non-recoverable ErrorCategory::Data.