[NancyKwongPerplexity]

Summary

Moves what a Native files:* request means to its Idempotency-Key out of space-api's files handlers into a new crate, pplx-space-native-files: RequestDigest, the per-operation request fingerprints (write, write_batch, mkdir, remove, rename) and the guest-key derivation (native-file-<digest>). A second server answering these routes, the data plane later in this stack, must derive byte-identical values, because a client retry can land on either server: the binding lives in public_idempotency_records and the sandbox replays writes and renames by the guest key.

The crate sits in crates/l40_api/native/files, the first crate in a layer directory. l20_domain would fit this PR alone, but the next PR adds request handling that needs pplx-space-common-grpc's status markers, and the one after it renders errors with axum. Both are fenced from l20_domain ([rules.layers.domain_fences]), so the crate starts in l40_api rather than moving later. The name avoids the banned catch-all tokens ([rules.crate_names]).

No behaviour change. The encodings move byte for byte, and the_sandbox_key_changes_with_the_request moves into the new crate with them. The crate's errors are a small FingerprintError, which space-api maps to the same internal error as before.

Validation

At this head (cbe1c6a1), on main at 0271415a:

• Rebased onto main at 0271415a. The only conflict was Cargo.lock: I took main's and let Cargo re-add this crate, which gives the same two lock entries as before. The commit is otherwise unchanged.
• cargo fmt --all -- --check, scripts/generate-env-contract.sh --check, scripts/check-rust-error-policy.sh and scripts/source_guards/env_test_hook_boundary.py pass.
• cargo clippy --locked --all-targets --all-features -p pplx-space-native-files -p pplx-space-api -- -D warnings passes.
• cargo check --workspace --all-targets --locked passes.
• cargo arch against main's baseline, renames and ARCHITECTURE.toml: 0 required findings and 112 advisory, the same findings as main.
• cargo nextest run --config 'profile.test.package.pplx-space-api.opt-level=0' --all-features --all-targets, one package at a time, with Postgres and Redis from the nextest setup scripts and SPACE_CI_REQUIRE_TEST_DEPS=1: pplx-space-api 2193 passed, 1 skipped; pplx-space-native-files 1 passed.
• Before the rebase, a throwaway test, not committed, compared each new fingerprint function with the inline digest chain it replaces, over every flag combination and with and without a user. Every value was equal.

scripts/check_no_inline_rust_test_modules.py fails on main already, on two pplx-space-observability files this change does not touch.

Severity and risk

Severity: does not apply (refactor).

Risk: a change in any fingerprint or guest-key encoding would turn retries into 409s for up to 10 minutes after a deploy, and could let a different request reach another one's guest record. The move is verbatim, and the existing idempotency tests in files/tests.rs pass unchanged. To roll back, revert this PR.

Stack

This is PR 1 of 5. Merge in order:

1. this PR
2. native-files: share Native file request handling #1363 (native-files-shared-handling)
3. public-api-http: make Native error rendering usable outside space-api #1364 (native-error-rendering-shared)
4. api: serve native file access RPCs on the activation listener #1365 (native-file-access-rpcs)
5. auth: let the data-plane identity call the native file access RPCs #1371 (data-plane-native-file-access-authz)

AI assistance: Claude Code drafted the code, tests, and this description; the author reviewed them.

[nikhilbirmiwal-pplx]

Checked that the write/write_batch/mkdir/remove/rename fingerprints, RequestDigest framing, and native-file- guest key move byte-for-byte into pplx-space-native-files, and FingerprintError maps to the same internal error.