Google

jyasskin@chromium.org

gen Dispatch Internet-Draft Web Packages provide a way to bundle up groups of web resources to transmit them together. These bundles can then be signed to establish their authenticity.

People would like to use content offline and in other situations where there isn’t a direct connection to the server where the content originates. However, it’s difficult to distribute and verify the authenticity of applications and content without a connection to the network. The W3C has addressed running applications offline with Service Workers (), but not the problem of distribution. We’ve started work on this problem in , but we suspect that the IETF may be the right place to standardize the overall format. More details can be found in that repository.

People with expensive or intermittent internet connections are used to sharing files via P2P links and shared SD cards. They should be able to install web applications they received this way. Installing a web application requires a TLS-type guarantee that it came from and can use data owned by a particular origin.

Verification of the origin of the content isn’t always necessary. For example, users currently share screenshots and MHTML documents with their peers, with no guarantee that the shared content is authentic. However, these formats have low fidelity (screenshots) and/or aren’t interoperable (MHTML). We’d like an interoperable format that lets both publishers and readers package such content for use in an untrusted mode.

CDNs want to re-publish other origins’ content so readers can access it more quickly or more privately. Currently, to attribute that content to the original origin, they need the full ability to publish arbitrary content under that origin’s name. There should be a way to let them attribute only the exact content that the original origin published.

WICG/webpackage#45

Publishers and readers should be able to generate a package once, and have it usable by all browsers.

The key words “MUST”, “MUST NOT”, “REQUIRED”, “SHALL”, “SHALL NOT”, “SHOULD”, “SHOULD NOT”, “RECOMMENDED”, “NOT RECOMMENDED”, “MAY”, and “OPTIONAL” in this document are to be interpreted as described in BCP 14 when, and only when, they appear in all capitals, as shown here.

This specification defines how conformant web package parsers convert a sequence of bytes into the semantics of a web package. It does not constrain how web package encoders produce such a package: although there are some guidelines in , encoders MAY produce any sequence of bytes that a conformant parser would parse into the intended semantics. In places, this specification says the parser “MAY return” some data. This indicates that the described data is complete enough that later parsing failures do not need to discard it. In places, this specification says the parser “MUST fail”. The parser MAY report these failures to its caller in any way, but MUST NOT return any data it has parsed so far that wasn’t mentioned in a “MAY return” statement. This specification creates local variables with the phrase “Let variable-name be …”. Use of a variable before it’s created is a defect in this specification.

The package is roughly a CBOR item with the following CDDL schema, but package parsers are required to successfully parse some byte strings that aren’t valid CBOR. For example, sections may have padding between them, or even overlap, as long as the embedded relative offsets cause the parsing algorithm in this specification to return data.

uint) }, sections: [ *$section ], length: uint, ; Total number of bytes in the package. magic2: h'F0 9F 8C 90 F0 9F 93 A6', ; 🌐📦 in UTF-8. ] ]]>

The parser MAY begin parsing at either the beginning or end of the byte string representing the package. Parsing from the end is useful when the package is embedded in another format such as a self-extracting executable, while parsing from the beginning is useful when loading from a stream.

To parse from the end, the parser MUST load the last 18 bytes as the following group in array context: CDDL doesn’t actually define how to use it as a schema to load CBOR data.

If the bytes don’t match this group or these two CBOR items don’t occupy exactly 18 bytes, parsing MUST fail. Otherwise, continue as if the byte length bytes before the end of the string were the beginning of the package, and the parser were a from the beginning parser.

If the first 10 bytes of the package are not “85 48 F0 9F 8C 90 F0 9F 93 A6” (the CBOR encoding of the 5-item array header and 8-byte bytestring header, followed by 🌐📦 in UTF-8), parsing MUST fail. Parse one CBOR item starting at the 11th byte of the package. If this does not match the CDDL

uint }, ]]>

or it is not a Canonical CBOR item (Section 3.9 of ), parsing MUST fail. Let sections-start be the offset of the byte after the section-offsets item. For example, if section-offsets were 52 bytes long, sections-start would be 63. This specification defines two section names: “indexed-content” and “manifest”. If section-offsets[“indexed-content”] is not present, parsing MUST fail. The parser MUST ignore unknown keys in the section-offsets map because new sections may be defined in future specifications. Do we need to mark critical section names? Let index be the result of parsing the bytes starting at offset sections-start + section-offsets[“indexed-content”] using the instructions in . If section-offsets[“manifest”] is present, let manifest be the result of parsing the bytes starting at offset sections-start + section-offsets[“manifest”] using the instructions in . The parser MAY return a semantic package consisting of index, and, if initialized, manifest. To parse each resource described within index, the parser MUST follow the instructions in .

The main content of a package is an index of HTTP requests pointing to HTTP responses. These request/response pairs hold the manifests of sub-packages and the resources in the package and all of its sub-packages. Both the requests and responses can appear in any order, usually chosen to optimize loading while the package is streamed. To parse the index, starting at offset index-start, the parser MUST do the following: If the byte at index-start is not 0x82 (the header for a 2-element array), the parser MUST fail. Load a CBOR item starting at index-start + 1 as the index array in the following CDDL:

If the item doesn’t match this CDDL, or it is not a Canonical CBOR item (Section 3.9 of ), the parser MUST fail. Let resources-start be the offset immediately after the index item. For example, if index-start were 75 and the index item were 105 bytes long, resources-start would be 75+1+105=181. (1 for the 0x82 array header.) Decode all of the resource-keys using , with an initially-empty dynamic table for each one. This spec has different security constraints from the ones that drove HPACK, so we may be able to do better with another compression format. The decoded resource-keys are header lists (, Section 1.3), ordered lists of name-value pairs. The parser MUST fail if any of the following is true: HPACK decoding encountered an error. Any resource-key’s first three headers are not named “:scheme”, “:authority”, and “:path”, in that order. Note that “:method” is intentionally omitted because only the GET method is meaningful. Any of the pseudo-headers’ values violates a requirement in Section 8.1.2.3 of . Any resource-key has a non-pseudo-header name that includes the “:” character or is not lower-case ascii (, Section 8.1.2). Any two decoded resource-keys are the same. Note that header lists with the same header fields in a different order are not the same. Increment all offsets by resources-start. Return the resulting index, an array of decoded-resource-key, adjusted-offset, and optional-length triples. The optional length field in the index entries is redundant with the length prefixes on the response-headers and body in the content, but it can be used to issue Range requests for responses that appear late in the content.

A package’s manifest contains some metadata for the package; hashes, used in , for all resources included in that package; and validity information for any sub-packages the package depends on. The manifest is signed, so that UAs can trust that it comes from its claimed origin. This section doesn’t describe a manifest (https://www.merriam-webster.com/dictionary/manifest#h3), so consider renaming it to something like “authenticity”. To parse a manifest starting at manifest-start, a parser MUST do the following: Load one CBOR item starting at manifest-start as a signed-manifest from the following CDDL:

[hash-value]}, ? subpackages: [* subpackage], } manifest-metadata = { date: time, origin: uri, * tstr => any, } ; From https://www.w3.org/TR/CSP3/#grammardef-hash-algorithm. hash-algorithm /= "sha256" / "sha384" / "sha512" ; Note that a hash value is not base64-encoded, unlike in CSP. hash-value = bstr ; X.509 format; see https://tools.ietf.org/html/rfc5280 certificate = bstr signature = { ; This is the index of the certificate within the certificates array to use ; to validate the signature. keyIndex: uint, signature: bstr, } ]]>

If the item doesn’t match the CDDL or it’s not a Canonical CBOR item (Section 3.9 of ), parsing MUST fail. Parse the elements of certificates as X.509 certificates within the profile. If any certificate fails to parse, parsing MUST fail. Let message be the concatenation of the following byte strings. This matches the format to avoid cross-protocol attacks when TLS certificates are used to sign manifests. A string that consists of octet 32 (0x20) repeated 64 times. A context string: the ASCII encoding of “Web Package Manifest”. A single 0 byte which serves as a separator. The bytes of the manifest CBOR item. Let signing-certificates be an empty array. For each element signature of signatures: Let certificate be certificates[signature[“keyIndex”]]. The parser MUST define a partial function from public key types to signing algorithms, with the following map as a subset: rsa_pss_sha256 as defined in Section 4.2.3 of ecdsa_secp256r1_sha256 as defined in Section 4.2.3 of ecdsa_secp384r1_sha384 as defined in Section 4.2.3 of Let signing-alg be the result of applying this function to the key type in certificate’s Subject Public Key Info. If the function is undefined on this input, the parser MUST continue to the next signature. Use signing-alg to verify that signature[“signature”] is message’s signature by certificate’s public key. If it’s not, the parser MUST continue to the next signature. Append certificate to signing-certificates. Note that failed signatures simply cause their certificate to be ignored, so that packagers can give new signature types to parsers that understand them. Let origin be manifest[“metadata”][“origin”]. Try to find a certificate in signing-certificates that has an identity (, Section 3.1) matching origin’s hostname, and that is trusted for serverAuth (, Section 4.2.1.12) using paths built from elements of certificates or any other certificates the parser is aware of. If no such certificate is found, and the package is not already trusted as received from origin’s hostname, for example because it was received over a TLS connection to that host, then parsing MUST fail. TODO: Process the subpackages item by fetching those manifests via the index, and checking their signatures and dates/hashes, recursively. The parsed manifest consists of the set of signing-certificates and the manifest CBOR item. The items in manifest[“metadata”] SHOULD be interpreted as described in the specification.

A sub-package is represented by a file looked up as a within the indexed-content section. The sub-package’s resources are not otherwise distinguished from the rest of the resources in the package. Sub-packages can form an arbitrarily-deep tree. There are three possible forms of dependencies on sub-packages, of which we allow two. Because a sub-package’s manifest is protected by its own signature, if the main package trusts the sub-package’s server, it could avoid specifying a version of the sub-package at all. However, this opens the main package up to downgrade attacks, where the sub-package is replaced by an older, vulnerable version, so we don’t allow this option.

hash-value}, ? notbefore: time, } ] ]]>

If the main package wants to load either the sub-package it was built with or any upgrade, it can specify the date of the original sub-package:

Constraining packages with their date makes it possible to link together sub-packages with common dependencies, even if the sub-packages were built at different times. If the main package wants to be certain it’s loading the exact version of a sub-package that it was built with, it can constrain sub-package with a hash of its manifest:

Note that because the sub-package may include sub-sub-packages by date, the top package may need to explicitly list those sub-sub-packages’ hashes in order to be completely constrained.

To parse the resource from a package corresponding to a header-list, a parser MUST do the following: Find the (resource-key, offset, length) triple in package’s index where resource-key is the same as header-list. If no such triple exists, the parser MUST fail. Parse one CBOR item starting at offset as the following CDDL:

If the item doesn’t match the CDDL or it’s not a Canonical CBOR item (Section 3.9 of ), parsing MUST fail. Decode the response-headers field using , with an initially-empty dynamic table. The decoded response-headers is a header list (, Section 1.3), an ordered list of name-value pairs. The parser MUST fail if any of the following is true: HPACK decoding encountered an error. The first header name within response-headers is not “:status”, or this pseudo-header’s value violates a requirement in Section 8.1.2.3 of . Any other header name includes the “:” character or is not lower-case ascii (, Section 8.1.2). The header-list contains any header names other than “:scheme”, “:authority”, “:path”, and either response-headers has no “vary” header (Section 7.1.4 of ) or these header names aren’t listed in it. Let origin be the Web Origin of header-list’s “:scheme” and “:authority” headers. Let resource-bytes be the result of encoding the array of [header-list, response-headers, body] as Canonical CBOR in the following CDDL schema: This step would be inside the manifest-only block, but then the code block is rendered out-of-order.

Note that this uses the decoded header fields, not the bytes originally included in the package. The hashed data differs from , which only hashes the body. Including the headers will usually prevent a package from relying on some of its contents being transferred as normal network responses, unless its author can guarantee the network won’t change or reorder the headers. If the package contains a manifest: TODO: Let origin-manifest be the signed manifest for origin, found by searching through manifest’s subpackages for a matching origin. Let alg be one of the hash-algorithms within origin-manifest. The parser SHOULD select the most collision-resistant hash algorithm. If the parser also implements , it SHOULD use the same order as its getPrioritizedHashFunction() implementation. If the digest of resource-bytes using alg does not appear in the origin-manifest’s resource-hashes[alg] array, the parser MUST fail. Return the (decoded response-headers, body) pair.

Packages SHOULD consist of a single Canonical CBOR item matching the webpackage CDDL rule in . Every resource’s hash SHOULD appear in every array within resource-hashes: otherwise the set of valid resources will depend on the parser’s choice of preferred hash algorithm.

Signature validation is difficult. Packages with a valid signature need to be invalidated when either the private key for any certificate in the signature’s validation chain is leaked, or a vulnerability is discovered in the package’s contents. Because packages are intended to be used offline, it’s impossible to inject a revocation check into the critical path of using the package, and even in online scenarios, such revocation checks don’t actually work. Instead, package consumers must check for a sufficiently recent set of validation files, consisting of OCSP responses and signed package version constraints, for example within the last 7-30 days. TODO: These version constraints aren’t designed yet. Relaxing the requirement to consult DNS when determining authority for an origin means that an attacker who possesses a valid certificate no longer needs to be on-path to redirect traffic to them; instead of modifying DNS, they need only convince the user to visit another Web site, in order to serve packages signed as the target. All subpackages that mention a particular origin need to be validated before loading resources from that origin. Otherwise, package A could include package B and an old, vulnerable version of package C that B also depends on. If B’s dependency isn’t checked before loading resources from C, A could compromise B.

IANA maintains the registry of Internet Media Types at https://www.iana.org/assignments/media-types. Type name: application Subtype name: package+cbor I suspect the mime type will need to be a bit longer: application/webpackage+cbor or similar. Required parameters: N/A Optional parameters: N/A Encoding considerations: binary Security considerations: See of this document. Interoperability considerations: N/A Published specification: This document Applications that use this media type: None yet, but it is expected that web browsers will use this format. Fragment identifier considerations: N/A Additional information: Deprecated alias names for this type: N/A Magic number(s): 85 48 F0 9F 8C 90 F0 9F 93 A6 File extension(s): .wpk Macintosh file type code(s): N/A Person & email address to contact for further information: See the Author’s Address section of this specification. Intended usage: COMMON Restrictions on usage: N/A Author: See the Author’s Address section of this specification. Change controller: The IESG iesg@ietf.org Provisional registration? (standards tree only): Not yet.

The Concise Binary Object Representation (CBOR) is a data format whose design goals include the possibility of extremely small code size, fairly small message size, and extensibility without the need for version negotiation. These design goals make it different from earlier binary serializations such as ASN.1 and MessagePack. This document proposes a notational convention to express CBOR data structures (RFC 7049). Its main goal is to provide an easy and unambiguous way to express structures for protocol messages and data formats that use CBOR. WHATWG In many standards track documents several words are used to signify the requirements in the specification. These words are often capitalized. This document defines these words as they should be interpreted in IETF documents. This document specifies an Internet Best Current Practices for the Internet Community, and requests discussion and suggestions for improvements. RFC 2119 specifies common key words that may be used in protocol specifications. This document aims to reduce the ambiguity by clarifying that only UPPERCASE usage of the key words have the defined special meanings. This specification defines HPACK, a compression format for efficiently representing HTTP header fields, to be used in HTTP/2. This specification describes an optimized expression of the semantics of the Hypertext Transfer Protocol (HTTP), referred to as HTTP version 2 (HTTP/2). HTTP/2 enables a more efficient use of network resources and a reduced perception of latency by introducing header field compression and allowing multiple concurrent exchanges on the same connection. It also introduces unsolicited push of representations from servers to clients.This specification is an alternative to, but does not obsolete, the HTTP/1.1 message syntax. HTTP's existing semantics remain unchanged. This memo profiles the X.509 v3 certificate and X.509 v2 certificate revocation list (CRL) for use in the Internet. An overview of this approach and model is provided as an introduction. The X.509 v3 certificate format is described in detail, with additional information regarding the format and semantics of Internet name forms. Standard certificate extensions are described and two Internet-specific extensions are defined. A set of required certificate extensions is specified. The X.509 v2 CRL format is described in detail along with standard and Internet-specific extensions. An algorithm for X.509 certification path validation is described. An ASN.1 module and examples are provided in the appendices. [STANDARDS-TRACK] This document specifies version 1.3 of the Transport Layer Security (TLS) protocol. TLS allows client/server applications to communicate over the Internet in a way that is designed to prevent eavesdropping, tampering, and message forgery. This memo describes how to use Transport Layer Security (TLS) to secure Hypertext Transfer Protocol (HTTP) connections over the Internet. This memo provides information for the Internet community. The Hypertext Transfer Protocol (HTTP) is a stateless \%application- level protocol for distributed, collaborative, hypertext information systems. This document defines the semantics of HTTP/1.1 messages, as expressed by request methods, request header fields, response status codes, and response header fields, along with the payload of messages (metadata and body content) and mechanisms for content negotiation. This document defines the concept of an "origin", which is often used as the scope of authority or privilege by user agents. Typically, user agents isolate content retrieved from different origins to prevent malicious web site operators from interfering with the operation of benign web sites. In addition to outlining the principles that underlie the concept of origin, this document details how to determine the origin of a URI and how to serialize an origin into a string. It also defines an HTTP header field, named "Origin", that indicates which origins are associated with an HTTP request. [STANDARDS-TRACK] This document specifies a protocol useful in determining the current status of a digital certificate without requiring Certificate Revocation Lists (CRLs). Additional mechanisms addressing PKIX operational requirements are specified in separate documents. This document obsoletes RFCs 2560 and 6277. It also updates RFC 5912. The Hypertext Transfer Protocol (HTTP) is a stateless application- level protocol for distributed, collaborative, hypertext information systems. This document defines range requests and the rules for constructing and combining responses to those requests. This document defines procedures for the specification and registration of media types for use in HTTP, MIME, and other Internet protocols. This memo documents an Internet Best Current Practice.

Thanks to Adam Langley and Ryan Sleevi for in-depth feedback about the security impact of this proposal.