Strictness & Failure semantics

Strata is strict by design.

Strictness is not an implementation detail or a configuration option. It is a core semantic property of the system.

When Strata encounters invalid input, it fails immediately and explicitly.

There is no recovery, no coercion, and no silent repair.

---

Why strictness exists

Strata exists to guarantee correctness at the byte level.

Any system that:

• hashes data

• signs data

• reaches consensus on data

• addresses data by content

cannot afford ambiguity.

Leniency introduces ambiguity. Ambiguity breaks determinism. Determinism is non-negotiable.

Strictness is how Strata protects its guarantees.

---

Failure is part of the contract

In Strata, failure is not an error state. It is an expected and meaningful outcome.

A failed operation communicates one thing clearly:

"This input violates the Strata contract."

That signal must never be weakened.

---

Encoding failure semantics

Encoding is authoritative and enforcing.

Encoding fails if:

• a value cannot be represented canonically

• an invariant would be violated

• an unsupported construct is encountered

• the value model is breached

Encoding never:

• guesses intent

• coerces types

• normalizes data silently

• repairs invalid input

If encoding fails, no bytes are produced.

This ensures that invalid data can never enter the system in canonical form.

---

Decoding failure semantics

Decoding is strict but interpretive.

Decoding fails if:

• an invalid tag is encountered

• input is truncated

• a varint overflows

• UTF-8 is invalid

• trailing bytes remain

• structural rules are violated

Decoding does not attempt recovery.

Malformed data is rejected immediately, at the exact offset where the violation occurs.

---

Error locality

All Strata failures are local and precise.

Errors report:

• the category of failure

• the exact byte offset (or source span)

• no speculative follow-up behavior

This makes Strata suitable for:

• auditing

• fuzzing

• security-sensitive environments

• protocol boundaries

---

No silent normalization

Strata never normalizes data silently.

Examples of behaviors Strata explicitly forbids:

• reordering maps on decode

• accepting non-canonical varints

• trimming trailing data

• fixing invalid UTF-8

• repairing malformed frames

If input is wrong, it stays wrong.

---

Failure vs flexibility

Many formats trade strictness for convenience:

• permissive decoders

• "best effort" parsing

• automatic repair

• backward-compatibility heuristics

Strata does not.

Flexibility in representation leads to ambiguity. Ambiguity breaks hashing and verification. Strata chooses correctness.

---

Security implications

Strict failure semantics:

• prevent downgrade attacks

• prevent canonicalization attacks

• prevent hash-collision ambiguity

• prevent parser differentials across languages

Every implementation fails the same way for the same input.

This is intentional.

---

Philosophy

Strata follows a simple rule:

If the system cannot prove correctness, it must refuse to operate.

Silence is dangerous. Explicit failure is safe.

---

Summary

• strictness is foundational, not optional

• encoding enforces invariants

• decoding rejects malformed input

• failure is explicit and precise

• no normalization, no recovery, no guessing

In Strata, failure is correctness asserting itself.