mini-static 0.17.0

A secure, async static file server with streaming, traversal protection, and connection limits.
Documentation

mini-static

A secure static file server: streaming responses, live reload, precompressed sidecars, external-tool CSS/JS bundling and minification, directory-index redirects. No templating, no framework — files in, HTTP responses out.

Status: published, actively developed. See DEV_PLAN.md for the roadmap.

[dependencies]
mini-static = "0.17"

Design

Server::new(root) canonicalizes root once at startup and serves everything below it. Every resolved path is checked against the canonicalized root before any file I/O happens — the canonicalization is the security boundary, not a pattern match on ...

Source folders and an output dir: optional source folders (with_source_folder) hold the inputs to the build pipelines — CSS that feeds a single bundle, JS that is minified per-file. A single output dir (with_output_dir, defaulting to the served root) receives the processed outputs. The two are disjoint by construction: a source folder that overlaps the output dir is rejected, and the output dir is never a watcher trigger — pipelines react to source folders only, so a pipeline's own output can never re-trigger its own rebuild.

Source-folder build pipelines

The server can build its own static output from with_source_folder(path) trees. When any source folder or a CSS/JS tool is configured, a startup build runs full_build; with with_live_reload() the source folders are watched and each change re-runs the appropriate step. Everything lives under the one output dir (with_output_dir, default the served root):

  • CSS and JS are each independently opt-in, via with_css_tool(tool, options) / with_js_tool(tool, options). With neither configured, source folders are watched (for live-reload) but nothing is transformed or copied to the output dir.
  • bundle/minify are independent toggles per language (CssOptions/JsOptions), all four combinations valid: passthrough copy, per-file minify (no @import/module resolution), bundle only (unminified, useful for debugging), or bundle + minify.
  • A source change re-runs only the affected step — a bundle-mode edit rebuilds the whole bundle, a per-file-mode edit rebuilds just that file — and then the reload broadcast is emitted after the output is written, so the browser reloads content that already exists.
  • Prune is opt-in and build-time only. with_prune_output() deletes the CSS bundle at startup when no CSS sources remain; it never runs during live-reload.
  • Asset folders (with_asset_folder(path)) are a third, simpler source kind. Every file under one — any extension, index.html, images, whatever — is mirrored byte-identical into the output dir, preserving its path relative to the asset folder. No CSS/JS tool involved, just a flat copy on startup and on every live-reload change. Use this for hand-authored static files that should live outside the served/output dir as source, the same separation the CSS/JS pipelines already have. Rejected if it overlaps the output dir or another registered source/asset folder, for the same feedback-loop reason with_source_folder is.
  • Server::build() runs every configured pipeline once and returns, no HTTP server involved. For deploy tooling that wants the output dir populated ahead of time — e.g. a one-shot cargo run --bin build_static step before baking a Docker image — mirroring a one-shot content builder's build() (such as mini_docs::Builder::build()) rather than needing to start-and-kill a live server just to get one build out of it.

Connection lifecycle (fixed from the prior iteration)

  • Header-read timeout. Every accepted connection now has a bounded time (default 30s) to send its request headers before the connection is dropped. The prior implementation had no timeout at all: a client that opened a socket and sent nothing held a connection-semaphore permit forever — 1024 idle sockets (the default max_connections), trivially cheap for an attacker, permanently stopped the server from accepting anyone else.
  • Ephemeral binds are loopback-only (127.0.0.1:0), matching mini-serve's fix and for the same reason: a test helper should never expose a real file server to the LAN.
  • A transient accept() error no longer ends the server. A sustained failure (e.g. the process is out of file descriptors) now degrades into periodic retries with exponential backoff instead of a single error silently ending the accept loop for good, or a naive retry busy-spinning at 100% CPU. Mirrors mini-serve's Backoff.
  • run()/run_ephemeral() return a ServerHandle alongside the port. The prior implementation had no way to stop a running server short of exiting the process — every test that started one leaked its background accept loop for the rest of the test binary's life, and an embedder had no way to stop serving at all. Calling handle.shutdown().await stops accepting new connections and waits for already-accepted connections to finish before returning; dropping the handle without calling it preserves the old fire-and-forget behavior.

Path traversal responses (fixed)

A blocked traversal attempt and a genuinely missing file both answer 404 not found. The prior implementation answered traversal attempts with a distinct 403 and a distinctive message — telling a prober exactly when they'd found the guard, and inviting iteration to map the filesystem by response code. The two responses are now byte-identical, so the distinction is not observable over the wire at all. Every file response also carries X-Content-Type-Options: nosniff — the server serves user-supplied directories, and content-sniffing a mislabeled file is a real vector for stored XSS.

The traversal pre-check matches path segments equal to .., not any substring containing .. — the prior substring check rejected legitimate filenames like jquery..min.js. canonicalize() + starts_with(root) remains the actual security boundary; the segment check is a cheap early rejection, not the guarantee.

HTTP correctness (fixed)

  • Method handling. Only GET/HEAD serve files; everything else gets 405 with Allow: GET, HEAD. The prior implementation served files for any method, including streaming the full body for HEAD requests before hyper silently dropped it on the wire — real disk I/O for a response nobody could see.
  • Content-Length on streamed responses. The file's length is known before streaming begins (metadata.len()) and is now sent on every full-file 200, not only on range responses — without it, clients fall back to chunked encoding and lose progress bars and cacheability by size.
  • Directory index redirects. Requesting /dir when /dir/index.html exists issues a 301 to /dir/ first, so relative links inside the served page resolve against the right base — the prior implementation served the index directly at /dir, silently breaking every relative link on the page.
  • Range requests get a full 200 with the whole body — Range and If-Range are ignored outright. This is the RFC 9110-correct answer for a server that doesn't implement 206 Partial Content (as opposed to an unhelpful 416), and it is honest about the fact that partial-content serving isn't built: nothing in the response claims range support, so no client negotiates for one.

Performance (fixed)

  • The server root is canonicalized once at startup; per-request resolution takes the already-canonical root as a documented precondition instead of re-canonicalizing (two syscalls plus an allocation) on every single request.
  • File responses stream to the client one chunk at a time via a Body impl backed by a reused BytesMut; each chunk is handed off via split_to(n).freeze() — no per-chunk zero-fill, no second copy of every byte read, and memory use stays bounded to one chunk per in-flight response regardless of file size.
  • Path resolution's blocking canonicalize() syscalls run on Tokio's blocking thread pool via spawn_blocking, not directly on the async worker thread handling the request — a slow filesystem lookup for one request no longer stalls every other task scheduled on that same worker thread.
  • Conditional-request support: the If-None-Match header is honored against the response's ETag, returning 304 Not Modified when the file hasn't changed — clients that revalidate get a fast, bodyless response instead of re-downloading the same content. If-Modified-Since is deliberately not implemented: an ETag already distinguishes representations that a whole-second mtime cannot (two writes inside the same second), so it is the validator to serve.

Filename decoding

Filenames are percent-decoded to raw bytes (not ASCII-only) and reassembled via OsStr::from_bytes on Unix, so non-ASCII filenames (é.png) are servable — the segment check and canonicalize() guard remain the authoritative boundary regardless of how the name was decoded.

Cache control and precompression

  • Cache-control default. Every 200/304 file response carries Cache-Control: no-cache — clients always revalidate against the ETag rather than caching blindly or getting no guidance at all.
  • Immutable assets. Server::with_immutable_assets(predicate) takes a Fn(&Path) -> bool; paths the predicate matches get Cache-Control: public, max-age=31536000, immutable instead of the default. Correct only for fingerprinted filenames (main.a1b2c3.js) where a content change always produces a new name — caching a mutable filename indefinitely would serve stale content to every client that already has it cached.
  • Precompressed sidecars. If a client's Accept-Encoding names br or gzip (br preferred when both are accepted and both sidecars exist) and a sibling <path>.br / <path>.gz exists next to the resolved file, its bytes are served instead with a matching Content-Encoding. Every file response carries Vary: Accept-Encoding so intermediate caches never serve the wrong variant to a differently-capable client, and the ETag reflects whichever variant was actually served — no compression dependency, a real bandwidth win for static sites that ship prebuilt .gz/.br files. The sidecar path is derived by appending an extension to the already-resolved, canonicalized path — never by re-resolving a modified request path — so it can't become a second traversal surface.

CSS/JS bundling and minification (external tools)

mini-static does not bundle or minify CSS/JS itself — it delegates to an external CLI tool you configure, and does not install or manage that tool. This keeps the core crate free of any particular bundler's dependency weight and release cadence, at the cost of an extra install step for the embedder.

  • Opt-in, per language, via a named preset. Server::with_css_tool(CssTool::LightningCss, options) and Server::with_js_tool(JsTool::Esbuild, options) — #[non_exhaustive] enums so more presets can be added later without breaking existing matches. Neither is configured by default.
  • bundle/minify are independent CssOptions/JsOptions toggles. All four combinations are valid per language: passthrough copy, per-file minify (mirrored 1:1 into the output dir, no @import/module resolution), bundle-only (unminified, concatenated/entry-resolved), or bundle + minify.
  • CSS bundling discovers every .css under the source folders and concatenates each file's (optionally @import-resolved, optionally minified) output in sorted path order — the same shape as before, just delegated per-file to the external tool.
  • JS bundling requires an explicit entry point (JsOptions::bundle_entry(path, output_name), validated to lie under a registered source folder at configuration time) — unlike CSS, a JS module graph has no well-defined "concatenate everything" meaning. Without a bundle entry, JS runs in per-file mode.
  • Installation is the embedder's responsibility. lightningcss (npm package lightningcss-cli) and/or esbuild (npm package esbuild) must be on PATH. Server::run_on checks this at startup — before the listener binds — and returns Err(StaticError::PipelineSetup) with an install hint if a configured tool's binary is missing, rather than silently serving unprocessed files.
  • Bounded and fails loudly. Every tool invocation has a 30-second timeout; a timeout, non-zero exit, or missing-binary error is reported with the tool's own stderr where available (see ToolError).
  • Per-file mode degrades to a raw copy on tool failure (a malformed source file, or the tool crashing on it) — logged loudly, not silently, so the server keeps serving and the browser stays in sync rather than 404ing.
  • A failed bundle rebuild leaves the previous good bundle in place — never a partial or corrupt file — and is logged; the next successful rebuild replaces it.
  • @import/module resolution is delegated entirely to the external tool — mini-static no longer enforces an import-root boundary or depth/file-count ceiling itself (the 30-second timeout is the bound in its place). This is an accepted trade-off: CSS/JS source folders are developer-authored build inputs, not request-time attacker input, unlike the HTTP path resolver above (which remains fully guarded). with_bundle_root still exists, but now purely as an extra watch target for triggering CSS rebuilds — not an @import traversal boundary.
  • *.min.css/*.min.js bypass per-file mode's minify step — already-minified files are mirrored as-is.
  • No request-time transformation. All processing is build-time only, via the source pipeline above — a subprocess call has unbounded latency and doesn't belong inside HTTP request handling. An embedder that wants CSS/JS served transformed must run the server with the appropriate source folder(s) and tool(s) configured.

Features

  • err — mini_err::Error integration for internal error responses.
  • log — request logging, including logged (not exposed) traversal attempts.

Non-goals

  • No directory listing UI — an index page is either present as a real file or the request 404s.
  • No on-the-fly transcoding or image resizing.
  • No built-in compression of arbitrary responses — see precompressed sidecar support above as the intended growth path instead.