feat(gateway): TAR response format

core/corehttp/gateway_handler.go

@@ -430,6 +430,10 @@ func (i *gatewayHandler) getOrHeadHandler(w http.ResponseWriter, r *http.Request
 		carVersion := formatParams["version"]
 		i.serveCAR(r.Context(), w, r, resolvedPath, contentPath, carVersion, begin)
 		return
+	case "application/x-tar":
+		logger.Debugw("serving tar file", "path", contentPath)
+		i.serveTAR(r.Context(), w, r, resolvedPath, contentPath, begin, logger)
+		return
 	default: // catch-all for unsuported application/vnd.*
 		err := fmt.Errorf("unsupported format %q", responseFormat)
 		webError(w, "failed respond with requested content type", err, http.StatusBadRequest)
@@ -842,9 +846,10 @@ func getEtag(r *http.Request, cid cid.Cid) string {
 	responseFormat, _, err := customResponseFormat(r)
 	if err == nil && responseFormat != "" {
 		// application/vnd.ipld.foo → foo
-		f := responseFormat[strings.LastIndex(responseFormat, ".")+1:]
-		// Etag: "cid.foo" (gives us nice compression together with Content-Disposition in block (raw) and car responses)
-		suffix = `.` + f + suffix
+		// application/x-bar → x-bar
+		shortFormat := responseFormat[strings.LastIndexAny(responseFormat, "/.")+1:]
+		// Etag: "cid.shortFmt" (gives us nice compression together with Content-Disposition in block (raw) and car responses)
+		suffix = `.` + shortFormat + suffix
 	}
 	// TODO: include selector suffix when https://github.com/ipfs/kubo/issues/8769 lands
 	return prefix + cid.String() + suffix
@@ -859,14 +864,17 @@ func customResponseFormat(r *http.Request) (mediaType string, params map[string]
 			return "application/vnd.ipld.raw", nil, nil
 		case "car":
 			return "application/vnd.ipld.car", nil, nil
+		case "tar":
+			return "application/x-tar", nil, nil
 		}
 	}
 	// Browsers and other user agents will send Accept header with generic types like:
 	// Accept:text/html,application/xhtml+xml,application/xml;q=0.9,image/avif,image/webp,*/*;q=0.8
-	// We only care about explciit, vendor-specific content-types.
+	// We only care about explicit, vendor-specific content-types.
 	for _, accept := range r.Header.Values("Accept") {
 		// respond to the very first ipld content type
-		if strings.HasPrefix(accept, "application/vnd.ipld") {
+		if strings.HasPrefix(accept, "application/vnd.ipld") ||
+			strings.HasPrefix(accept, "application/x-tar") {
 			mediatype, params, err := mime.ParseMediaType(accept)
 			if err != nil {
 				return "", nil, err

core/corehttp/gateway_handler_tar.go

@@ -0,0 +1,92 @@
+package corehttp
+
+import (
+	"context"
+	"html"
+	"net/http"
+	"time"
+
+	files "github.com/ipfs/go-ipfs-files"
+	ipath "github.com/ipfs/interface-go-ipfs-core/path"
+	"github.com/ipfs/kubo/tracing"
+	"go.opentelemetry.io/otel/attribute"
+	"go.opentelemetry.io/otel/trace"
+	"go.uber.org/zap"
+)
+
+var unixEpochTime = time.Unix(0, 0)
+
+func (i *gatewayHandler) serveTAR(ctx context.Context, w http.ResponseWriter, r *http.Request, resolvedPath ipath.Resolved, contentPath ipath.Path, begin time.Time, logger *zap.SugaredLogger) {
+	ctx, span := tracing.Span(ctx, "Gateway", "ServeTAR", trace.WithAttributes(attribute.String("path", resolvedPath.String())))
+	defer span.End()
+
+	ctx, cancel := context.WithCancel(ctx)
+	defer cancel()
+
+	// Get Unixfs file
+	file, err := i.api.Unixfs().Get(ctx, resolvedPath)
+	if err != nil {
+		webError(w, "ipfs cat "+html.EscapeString(contentPath.String()), err, http.StatusBadRequest)
+		return
+	}
+	defer file.Close()
+
+	rootCid := resolvedPath.Cid()
+
+	// Set Cache-Control and read optional Last-Modified time
+	modtime := addCacheControlHeaders(w, r, contentPath, rootCid)
+
+	// Weak Etag W/ because we can't guarantee byte-for-byte identical
+	// responses, but still want to benefit from HTTP Caching. Two TAR
+	// responses for the same CID will be logically equivalent,
+	// but when TAR is streamed, then in theory, files and directories
+	// may arrive in different order (depends on TAR lib and filesystem/inodes).
+	etag := `W/` + getEtag(r, rootCid)
+	w.Header().Set("Etag", etag)
+
+	// Finish early if Etag match
+	if r.Header.Get("If-None-Match") == etag {
+		w.WriteHeader(http.StatusNotModified)
+		return
+	}
+
+	// Set Content-Disposition
+	var name string
+	if urlFilename := r.URL.Query().Get("filename"); urlFilename != "" {
+		name = urlFilename
+	} else {
+		name = rootCid.String() + ".tar"
+	}
+	setContentDispositionHeader(w, name, "attachment")
+
+	// Construct the TAR writer
+	tarw, err := files.NewTarWriter(w)
+	if err != nil {
+		webError(w, "could not build tar writer", err, http.StatusInternalServerError)
+		return
+	}
+	defer tarw.Close()
+
+	// Sets correct Last-Modified header. This code is borrowed from the standard
+	// library (net/http/server.go) as we cannot use serveFile without throwing the entire
+	// TAR into the memory first.
+	if !(modtime.IsZero() || modtime.Equal(unixEpochTime)) {
+		w.Header().Set("Last-Modified", modtime.UTC().Format(http.TimeFormat))
+	}
+
+	w.Header().Set("Content-Type", "application/x-tar")
+	w.Header().Set("X-Content-Type-Options", "nosniff") // no funny business in the browsers :^)
+
+	// The TAR has a top-level directory (or file) named by the CID.
+	if err := tarw.WriteFile(file, rootCid.String()); err != nil {
+		w.Header().Set("X-Stream-Error", err.Error())
+		// Trailer headers do not work in web browsers
+		// (see https://github.com/mdn/browser-compat-data/issues/14703)
+		// and we have limited options around error handling in browser contexts.
+		// To improve UX/DX, we finish response stream with error message, allowing client to
+		// (1) detect error by having corrupted TAR
+		// (2) be able to reason what went wrong by instecting the tail of TAR stream
+		_, _ = w.Write([]byte(err.Error()))
+		return
+	}
+}

docs/changelogs/v0.17.md

@@ -9,15 +9,34 @@ Below is an outline of all that is in this release, so you get a sense of all th
 - [Kubo changelog v0.17](#kubo-changelog-v017)
   - [v0.17.0](#v0170)
     - [Overview](#overview)
+    - [TOC](#toc)
     - [🔦 Highlights](#-highlights)
+      - [TAR Response Format on Gateways](#tar-response-format-on-gateways)
     - [Changelog](#changelog)
     - [Contributors](#contributors)
 
-
 ### 🔦 Highlights
 
 <!-- TODO -->
 
+#### TAR Response Format on Gateways
+
+Implemented [IPIP-288](https://github.com/ipfs/specs/pull/288) which adds
+support for requesting deserialized UnixFS directory as a TAR stream.
+
+HTTP clients can request TAR response by passing the `?format=tar` URL
+parameter, or setting `Accept: application/x-tar` HTTP header:
+
+```console
+$ export DIR_CID=bafybeigccimv3zqm5g4jt363faybagywkvqbrismoquogimy7kvz2sj7sq
+$ curl -H "Accept: application/x-tar" "http://127.0.0.1:8080/ipfs/$DIR_CID" > dir.tar
+$ curl "http://127.0.0.1:8080/ipfs/$DIR_CID?format=tar" | tar xv
+bafybeigccimv3zqm5g4jt363faybagywkvqbrismoquogimy7kvz2sj7sq
+bafybeigccimv3zqm5g4jt363faybagywkvqbrismoquogimy7kvz2sj7sq/1 - Barrel - Part 1 - alt.txt
+bafybeigccimv3zqm5g4jt363faybagywkvqbrismoquogimy7kvz2sj7sq/1 - Barrel - Part 1 - transcript.txt
+bafybeigccimv3zqm5g4jt363faybagywkvqbrismoquogimy7kvz2sj7sq/1 - Barrel - Part 1.png
+```
+
 ### Changelog
 
 <!-- TODO -->

docs/examples/kubo-as-a-library/go.mod

@@ -7,7 +7,7 @@ go 1.17
 replace github.com/ipfs/kubo => ./../../..
 
 require (
-	github.com/ipfs/go-ipfs-files v0.1.1
+	github.com/ipfs/go-ipfs-files v0.2.0
 	github.com/ipfs/interface-go-ipfs-core v0.7.0
 	github.com/ipfs/kubo v0.14.0-rc1
 	github.com/libp2p/go-libp2p v0.23.2

docs/examples/kubo-as-a-library/go.sum

@@ -559,8 +559,8 @@ github.com/ipfs/go-ipfs-exchange-offline v0.3.0 h1:c/Dg8GDPzixGd0MC8Jh6mjOwU57uY
 github.com/ipfs/go-ipfs-exchange-offline v0.3.0/go.mod h1:MOdJ9DChbb5u37M1IcbrRB02e++Z7521fMxqCNRrz9s=
 github.com/ipfs/go-ipfs-files v0.0.3/go.mod h1:INEFm0LL2LWXBhNJ2PMIIb2w45hpXgPjNoE7yA8Y1d4=
 github.com/ipfs/go-ipfs-files v0.0.8/go.mod h1:wiN/jSG8FKyk7N0WyctKSvq3ljIa2NNTiZB55kpTdOs=
-github.com/ipfs/go-ipfs-files v0.1.1 h1:/MbEowmpLo9PJTEQk16m9rKzUHjeP4KRU9nWJyJO324=
-github.com/ipfs/go-ipfs-files v0.1.1/go.mod h1:8xkIrMWH+Y5P7HvJ4Yc5XWwIW2e52dyXUiC0tZyjDbM=
+github.com/ipfs/go-ipfs-files v0.2.0 h1:z6MCYHQSZpDWpUSK59Kf0ajP1fi4gLCf6fIulVsp8A8=
+github.com/ipfs/go-ipfs-files v0.2.0/go.mod h1:vT7uaQfIsprKktzbTPLnIsd+NGw9ZbYwSq0g3N74u0M=
 github.com/ipfs/go-ipfs-keystore v0.0.2 h1:Fa9xg9IFD1VbiZtrNLzsD0GuELVHUFXCWF64kCPfEXU=
 github.com/ipfs/go-ipfs-keystore v0.0.2/go.mod h1:H49tRmibOEs7gLMgbOsjC4dqh1u5e0R/SWuc2ScfgSo=
 github.com/ipfs/go-ipfs-pinner v0.2.1 h1:kw9hiqh2p8TatILYZ3WAfQQABby7SQARdrdA+5Z5QfY=

go.mod

@@ -37,7 +37,7 @@ require (
 	github.com/ipfs/go-ipfs-cmds v0.8.1
 	github.com/ipfs/go-ipfs-exchange-interface v0.2.0
 	github.com/ipfs/go-ipfs-exchange-offline v0.3.0
-	github.com/ipfs/go-ipfs-files v0.1.1
+	github.com/ipfs/go-ipfs-files v0.2.0
 	github.com/ipfs/go-ipfs-keystore v0.0.2
 	github.com/ipfs/go-ipfs-pinner v0.2.1
 	github.com/ipfs/go-ipfs-posinfo v0.0.1

go.sum

@@ -552,8 +552,8 @@ github.com/ipfs/go-ipfs-exchange-offline v0.3.0 h1:c/Dg8GDPzixGd0MC8Jh6mjOwU57uY
 github.com/ipfs/go-ipfs-exchange-offline v0.3.0/go.mod h1:MOdJ9DChbb5u37M1IcbrRB02e++Z7521fMxqCNRrz9s=
 github.com/ipfs/go-ipfs-files v0.0.3/go.mod h1:INEFm0LL2LWXBhNJ2PMIIb2w45hpXgPjNoE7yA8Y1d4=
 github.com/ipfs/go-ipfs-files v0.0.8/go.mod h1:wiN/jSG8FKyk7N0WyctKSvq3ljIa2NNTiZB55kpTdOs=
-github.com/ipfs/go-ipfs-files v0.1.1 h1:/MbEowmpLo9PJTEQk16m9rKzUHjeP4KRU9nWJyJO324=
-github.com/ipfs/go-ipfs-files v0.1.1/go.mod h1:8xkIrMWH+Y5P7HvJ4Yc5XWwIW2e52dyXUiC0tZyjDbM=
+github.com/ipfs/go-ipfs-files v0.2.0 h1:z6MCYHQSZpDWpUSK59Kf0ajP1fi4gLCf6fIulVsp8A8=
+github.com/ipfs/go-ipfs-files v0.2.0/go.mod h1:vT7uaQfIsprKktzbTPLnIsd+NGw9ZbYwSq0g3N74u0M=
 github.com/ipfs/go-ipfs-keystore v0.0.2 h1:Fa9xg9IFD1VbiZtrNLzsD0GuELVHUFXCWF64kCPfEXU=
 github.com/ipfs/go-ipfs-keystore v0.0.2/go.mod h1:H49tRmibOEs7gLMgbOsjC4dqh1u5e0R/SWuc2ScfgSo=
 github.com/ipfs/go-ipfs-pinner v0.2.1 h1:kw9hiqh2p8TatILYZ3WAfQQABby7SQARdrdA+5Z5QfY=

test/sharness/t0122-gateway-tar.sh

@@ -0,0 +1,91 @@
+#!/usr/bin/env bash
+
+test_description="Test HTTP Gateway TAR (application/x-tar) Support"
+
+. lib/test-lib.sh
+
+test_init_ipfs
+test_launch_ipfs_daemon_without_network
+
+OUTSIDE_ROOT_CID="bafybeicaj7kvxpcv4neaqzwhrqqmdstu4dhrwfpknrgebq6nzcecfucvyu"
+INSIDE_ROOT_CID="bafybeibfevfxlvxp5vxobr5oapczpf7resxnleb7tkqmdorc4gl5cdva3y"
+
+test_expect_success "Add the test directory" '
+  mkdir -p rootDir/ipfs &&
+  mkdir -p rootDir/ipns &&
+  mkdir -p rootDir/api &&
+  mkdir -p rootDir/ą/ę &&
+  echo "I am a txt file on path with utf8" > rootDir/ą/ę/file-źł.txt &&
+  echo "I am a txt file in confusing /api dir" > rootDir/api/file.txt &&
+  echo "I am a txt file in confusing /ipfs dir" > rootDir/ipfs/file.txt &&
+  echo "I am a txt file in confusing /ipns dir" > rootDir/ipns/file.txt &&
+  DIR_CID=$(ipfs add -Qr --cid-version 1 rootDir) &&
+  FILE_CID=$(ipfs files stat --enc=json /ipfs/$DIR_CID/ą/ę/file-źł.txt | jq -r .Hash) &&
+  FILE_SIZE=$(ipfs files stat --enc=json /ipfs/$DIR_CID/ą/ę/file-źł.txt | jq -r .Size)
+  echo "$FILE_CID / $FILE_SIZE"
+'
+
+test_expect_success "GET TAR with format=tar and extract" '
+  curl "http://127.0.0.1:$GWAY_PORT/ipfs/$FILE_CID?format=tar" | tar -x
+'
+
+test_expect_success "GET TAR with 'Accept: application/x-tar' and extract" '
+  curl -H "Accept: application/x-tar" "http://127.0.0.1:$GWAY_PORT/ipfs/$FILE_CID" | tar -x
+'
+
+test_expect_success "GET TAR with format=tar has expected Content-Type" '
+  curl -sD - "http://127.0.0.1:$GWAY_PORT/ipfs/$FILE_CID?format=tar" > curl_output_filename 2>&1 &&
+  test_should_contain "Content-Disposition: attachment;" curl_output_filename &&
+  test_should_contain "Etag: W/\"$FILE_CID.x-tar" curl_output_filename &&
+  test_should_contain "Content-Type: application/x-tar" curl_output_filename
+'
+
+test_expect_success "GET TAR with 'Accept: application/x-tar' has expected Content-Type" '
+  curl -sD - -H "Accept: application/x-tar" "http://127.0.0.1:$GWAY_PORT/ipfs/$FILE_CID" > curl_output_filename 2>&1 &&
+  test_should_contain "Content-Disposition: attachment;" curl_output_filename &&
+  test_should_contain "Etag: W/\"$FILE_CID.x-tar" curl_output_filename &&
+  test_should_contain "Content-Type: application/x-tar" curl_output_filename
+'
+
+test_expect_success "GET TAR has expected root file" '
+  rm -rf outputDir && mkdir outputDir &&
+  curl "http://127.0.0.1:$GWAY_PORT/ipfs/$FILE_CID?format=tar" | tar -x -C outputDir &&
+  test -f "outputDir/$FILE_CID" &&
+  echo "I am a txt file on path with utf8" > expected &&
+  test_cmp expected outputDir/$FILE_CID
+'
+
+test_expect_success "GET TAR has expected root directory" '
+  rm -rf outputDir && mkdir outputDir &&
+  curl "http://127.0.0.1:$GWAY_PORT/ipfs/$DIR_CID?format=tar" | tar -x -C outputDir &&
+  test -d "outputDir/$DIR_CID" &&
+  echo "I am a txt file on path with utf8" > expected &&
+  test_cmp expected outputDir/$DIR_CID/ą/ę/file-źł.txt
+'
+
+test_expect_success "GET TAR with explicit ?filename= succeeds with modified Content-Disposition header" "
+  curl -fo actual -D actual_headers 'http://127.0.0.1:$GWAY_PORT/ipfs/$DIR_CID?filename=testтест.tar&format=tar' &&
+  grep -F 'Content-Disposition: attachment; filename=\"test____.tar\"; filename*=UTF-8'\'\''test%D1%82%D0%B5%D1%81%D1%82.tar' actual_headers
+"
+
+test_expect_success "Add CARs with relative paths to test with" '
+  ipfs dag import ../t0122-gateway-tar-data/outside-root.car > import_output &&
+  test_should_contain $OUTSIDE_ROOT_CID import_output &&
+  ipfs dag import ../t0122-gateway-tar-data/inside-root.car > import_output &&
+  test_should_contain $INSIDE_ROOT_CID import_output
+'
+
+test_expect_success "GET TAR with relative paths outside root fails" '
+  curl -o - "http://127.0.0.1:$GWAY_PORT/ipfs/$OUTSIDE_ROOT_CID?format=tar" > curl_output_filename &&
+  test_should_contain "relative UnixFS paths outside the root are now allowed" curl_output_filename
+'
+
+test_expect_success "GET TAR with relative paths inside root works" '
+  rm -rf outputDir && mkdir outputDir &&
+  curl "http://127.0.0.1:$GWAY_PORT/ipfs/$INSIDE_ROOT_CID?format=tar" | tar -x -C outputDir &&
+  test -f outputDir/$INSIDE_ROOT_CID/foobar/file
+'
+
+test_kill_ipfs_daemon
+
+test_done