Akamai

mnot@mnot.net http://www.mnot.net/

Mozilla

mcmanus@ducksong.com https://mozillians.org/u/pmcmanus/

General Internet-Draft This document specifies “alternative services” for HTTP, which allow an origin’s resources to be authoritatively available at a separate network location, possibly accessed with a different protocol configuration.

HTTP conflates the identification of resources with their location. In other words, http:// (and https://) URLs are used to both name and find things to interact with. In some cases, it is desirable to separate these aspects; to be able to keep the same identifier for a resource, but interact with it using a different location on the network. For example: An origin server might wish to redirect a client to an alternative when it needs to go down for maintenance, or it has found an alternative in a location that is more local to the client. An origin server might wish to offer access to its resources using a new protocol (such as HTTP/2 ) or one using improved security (such as TLS {{RFC5246}). An origin server might wish to segment its clients into groups of capabilities, such as those supporting SNI (see ) and those not supporting it, for operational purposes. This specification defines a new concept in HTTP, “Alternative Services”, that allows a resource to nominate additional means of interacting with it on the network. It defines a general framework for this in , along with a specific mechanism for discovering them using HTTP headers in . It also introduces a new status code in , so that origin servers (or their nominated alternatives) can indicate that they are not authoritative for a given origin, in cases where the wrong location is used.

The key words “MUST”, “MUST NOT”, “REQUIRED”, “SHALL”, “SHALL NOT”, “SHOULD”, “SHOULD NOT”, “RECOMMENDED”, “MAY”, and “OPTIONAL” in this document are to be interpreted as described in . This document uses the Augmented BNF defined in along with the “OWS”, “DIGIT”, “parameter”, “port” and “delta-second” rules from , and uses the “#rule” extension defined in Section 7 of that document.

This specification defines a new concept in HTTP, the “alternative service.” When an origin (see ) has resources are accessible through a different protocol / host / port combination, it is said to have an alternative service. An alternative service can be used to interact with the resources on an origin server at a separate location on the network, possibly using a different protocol configuration. Alternative services are considered authoritative for an origin’s resources, in the sense of Section 9.1. For example, an origin:

might declare that its resources are also accessible at the alternative service:

By their nature, alternative services are explicitly at the granularity of an origin; i.e., they cannot be selectively applied to resources within an origin. Alternative services do not replace or change the origin for any given resource; in general, they are not visible to the software “above” the access mechanism. The alternative service is essentially alternative routing information that can also be used to reach the origin in the same way that DNS CNAME or SRV records define routing information at the name resolution level. Each origin maps to a set of these routes - the default route is derived from origin itself and the other routes are introduced based on alternative-protocol information. Furthermore, it is important to note that the first member of an alternative service tuple is different from the “scheme” component of an origin; it is more specific, identifying not only the major version of the protocol being used, but potentially communication options for that protocol. This means that clients using an alternative service will change the host, port and protocol that they are using to fetch resources, but these changes MUST NOT be propagated to the application that is using HTTP; from that standpoint, the URI being accessed and all information derived from it (scheme, host, port) are the same as before. Importantly, this includes its security context; in particular, when TLS is in use, the alternative server will need to present a certificate for the origin’s host name, not that of the alternative. Likewise, the Host header is still derived from the origin, not the alternative service (just as it would if a CNAME were being used). The changes MAY, however, be made visible in debugging tools, consoles, etc. Formally, an alternative service is identified by the combination of: An ALPN protocol, as per A host, as per A port, as per Additionally, each alternative service MUST have: A freshness lifetime, expressed in seconds; see There are many ways that a client could discover the alternative service(s) associated with an origin.

Clients MUST NOT use alternative services with a host other than the origin’s without strong server authentication; this mitigates the attack described in . One way to achieve this is for the alternative to use TLS with a certificate that is valid for that origin. For example, if the origin’s host is “www.example.com” and an alternative is offered on “other.example.com” with the “h2” protocol, and the certificate offered is valid for “www.example.com”, the client can use the alternative. However, if “other.example.com” is offered with the “h2c” protocol, the client cannot use it, because there is no mechanism in that protocol to establish strong server authentication. Furthermore, this means that the HTTP Host header and the SNI information provided in TLS by the client will be that of the origin, not the alternative.

Mechanisms for discovering alternative services can associate a freshness lifetime with them; for example, the Alt-Svc header field uses the “ma” parameter. Clients MAY choose to use an alternative service instead of the origin at any time when it is considered fresh; see for specific recommendations. Clients with existing connections to alternative services are not required to fall back to the origin when its freshness lifetime ends; i.e., the caching mechanism is intended for limiting how long an alternative service can be used for establishing new requests, not limiting the use of existing ones. To mitigate risks associated with caching compromised values (see for details), user agents SHOULD examine cached alternative services when they detect a change in network configuration, and remove any that could be compromised (for example, those whose association with the trust root is questionable). UAs that do not have a means of detecting network changes SHOULD place an upper bound on their lifetime.

A client must only use a TLS-based alternative service if the client also supports TLS Server Name Indication (SNI) . This supports the conservation of IP addresses on the alternative service host.

By their nature, alternative services are optional; clients are not required to use them. However, it is advantageous for clients to behave in a predictable way when they are used by servers (e.g., for load balancing). Therefore, if a client becomes aware of an alternative service, the client SHOULD use that alternative service for all requests to the associated origin as soon as it is available, provided that the security properties of the alternative service protocol are desirable, as compared to the existing connection. The client is not required to block requests; the origin’s connection can be used until the alternative connection is established. However, if the security properties of the existing connection are weak (e.g. cleartext HTTP/1.1) then it might make sense to block until the new connection is fully available in order to avoid information leakage. Furthermore, if the connection to the alternative service fails or is unresponsive, the client MAY fall back to using the origin. Note, however, that this could be the basis of a downgrade attack, thus losing any enhanced security properties of the alternative service.

A HTTP(S) origin server can advertise the availability of alternative services (see ) to clients by adding an Alt-Svc header field to responses.

protocol-id <"> "=" port protocol-id = ]]>

For example:

This indicates that the “http2” protocol on the same host using the indicated port (in this case, 8000). Alt-Svc MAY occur in any HTTP response message, regardless of the status code. Alt-Svc does not allow advertisement of alternative services on other hosts, to protect against various header-based attacks. It can, however, have multiple values:

The value(s) advertised by Alt-Svc can be used by clients to open a new connection to one or more alternative services immediately, or simultaneously with subsequent requests on the same connection. Intermediaries MUST NOT change or append Alt-Svc values. Finally, note that while it may be technically possible to put content other than printable ASCII in a HTTP header, some implementations only support ASCII (or a superset of it) in header field values. Therefore, this field SHOULD NOT be used to convey protocol identifiers that are not printable ASCII, or those that contain quote characters.

When an alternative service is advertised using Alt-Svc, it is considered fresh for 24 hours from generation of the message. This can be modified with the ‘ma’ (max-age) parameter;

which indicates the number of seconds since the response was generated the alternative service is considered fresh for.

See Section 4.2.3 for details of determining response age. For example, a response:

indicates that an alternative service is available and usable for the next 60 seconds. However, the response has already been cached for 30 seconds (as per the Age header field value), so therefore the alternative service is only fresh for the 30 seconds from when this response was received, minus estimated transit time. When an Alt-Svc response header is received from an origin, its value invalidates and replaces all cached alternative services for that origin. See for general requirements on caching alternative services. Note that the freshness lifetime for HTTP caching (here, 600 seconds) does not affect caching of Alt-Svc values.

The 4NN (Not Authoritative) status code indicates that the current origin server (usually, but not always an alternative service; see ) is not authoritative for the requested resource, in the sense of , Section 9.1. Clients receiving 4NN (Not Authoritative) from an alternative service MUST remove the corresponding entry from its alternative service cache (see ) for that origin. Regardless of the idempotency of the request method, they MAY retry the request, either at another alternative server, or at the origin. 4NN (Not Authoritative) MAY carry an Alt-Svc header field. This status code MUST NOT be generated by proxies. A 4NN response is cacheable by default; i.e., unless otherwise indicated by the method definition or explicit cache controls (see Section 4.2.2 of ).

This document registers Alt-Svc in the Permanent Message Header Registry . Header Field Name: Alt-Svc Application Protocol: http Status: standard Author/Change Controller: IETF Specification Document: [this document] Related Information:

This document registers the 4NN (Not Authoritative) HTTP Status code . Status Code: 4NN Short Description: Not Authoritative Specification: [this document],

Identified security considerations should be enumerated in the appropriate documents depending on which proposals are accepted. Those listed below are generic to all uses of alternative services; more specific ones might be necessary.

Using an alternative service implies accessing an origin’s resources on an alternative port, at a minimum. An attacker that can inject alternative services and listen at the advertised port is therefore able to hijack an origin. For example, an attacker that can add HTTP response header fields can redirect traffic to a different port on the same host using the Alt-Svc header field; if that port is under the attacker’s control, they can thus masquerade as the HTTP server. This risk can be mitigated by restricting the ability to advertise alternative services, and restricting who can open a port for listening on that host.

When the host is changed due to the use of an alternative service, it presents an opportunity for attackers to hijack communication to an origin. For example, if an attacker can convince a user agent to send all traffic for “innocent.example.org” to “evil.example.com” by successfully associating it as an alternative service, they can masquerade as that origin. This can be done locally (see mitigations above) or remotely (e.g., by an intermediary as a man-in-the-middle attack). This is the reason for the requirement in that any alternative service with a host different to the origin’s be strongly authenticated with the origin’s identity; i.e., presenting a certificate for the origin proves that the alternative service is authorized to serve traffic for the origin. However, this authorization is only as strong as the method used to authenticate the alternative service. In particular, there are well-known exploits to make an attacker’s certificate appear as legitimate. Alternative services could be used to persist such an attack; for example, an intermediary could man-in-the-middle TLS-protected communication to a target, and then direct all traffic to an alternative service with a large freshness lifetime, so that the user agent still directs traffic to the attacker even when not using the intermediary. As a result, there is a requirement in to examine cached alternative services when a network change is detected.

When the ALPN protocol is changed due to the use of an alternative service, the security properties of the new connection to the origin can be different from that of the “normal” connection to the origin, because the protocol identifier itself implies this. For example, if a “https://” URI had a protocol advertised that does not use some form of end-to-end encryption (most likely, TLS), it violates the expectations for security that the URI scheme implies. Therefore, clients cannot blindly use alternative services, but instead evaluate the option(s) presented to assure that security requirements and expectations (of specifications, implementations and end users) are met.

Harvard University

1350 Mass. Ave. Cambridge MA 02138 - +1 617 495 3864 sob@harvard.edu

General keyword 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. Authors who follow these guidelines should incorporate this phrase near the beginning of their document: The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119. Note that the force of these words is modified by the requirement level of the document in which they are used. World Wide Web Consortium

Massachusetts Institute of Technology 77 Massachusetts Avenue Cambridge MA 02139 USA +1-617-253-5702 +1-617-258-5999 timbl@w3.org http://www.w3.org/People/Berners-Lee/

Day Software

5251 California Ave., Suite 110 Irvine CA 92617 USA +1-949-679-2960 +1-949-679-2972 fielding@gbiv.com http://roy.gbiv.com/

Adobe Systems Incorporated

345 Park Ave San Jose CA 95110 USA +1-408-536-3024 LMM@acm.org http://larry.masinter.net/

Applications uniform resource identifier URI URL URN WWW resource A Uniform Resource Identifier (URI) is a compact sequence of characters that identifies an abstract or physical resource. This specification defines the generic URI syntax and a process for resolving URI references that might be in relative form, along with guidelines and security considerations for the use of URIs on the Internet. The URI syntax defines a grammar that is a superset of all valid URIs, allowing an implementation to parse the common components of a URI reference without knowing the scheme-specific requirements of every possible identifier. This specification does not define a generative grammar for URIs; that task is performed by the individual specifications of each URI scheme. Internet technical specifications often need to define a formal syntax. Over the years, a modified version of Backus-Naur Form (BNF), called Augmented BNF (ABNF), has been popular among many Internet specifications. The current specification documents ABNF. It balances compactness and simplicity with reasonable representational power. The differences between standard BNF and ABNF involve naming rules, repetition, alternatives, order-independence, and value ranges. This specification also supplies additional rule definitions and encoding for a core lexical analyzer of the type common to several Internet specifications. [STANDARDS-TRACK] This document provides specifications for existing TLS extensions. It is a companion document for RFC 5246, "The Transport Layer Security (TLS) Protocol Version 1.2". The extensions specified are server_name, max_fragment_length, client_certificate_url, trusted_ca_keys, truncated_hmac, and status_request. [STANDARDS-TRACK] 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 describes a Transport Layer Security (TLS) extension for application layer protocol negotiation within the TLS handshake. For instances in which the TLS connection is established over a well known TCP or UDP port not associated with the desired application layer protocol, this extension allows the application layer to negotiate which protocol will be used within the TLS connection. The Hypertext Transfer Protocol (HTTP) is a stateless application- level protocol for distributed, collaborative, hypertext information systems. This document provides an overview of HTTP architecture and its associated terminology, defines the "http" and "https" Uniform Resource Identifier (URI) schemes, defines the HTTP/1.1 message syntax and parsing requirements, and describes related security concerns for implementations. The Hypertext Transfer Protocol (HTTP) is a stateless application- level protocol for distributed, collaborative, hypertext information systems. This document defines HTTP caches and the associated header fields that control cache behavior or indicate cacheable response messages. This specification defines registration procedures for the message header fields used by Internet mail, HTTP, Netnews and other applications. This document specifies an Internet Best Current Practices for the Internet Community, and requests discussion and suggestions for improvements. This document specifies Version 1.2 of the Transport Layer Security (TLS) protocol. The TLS protocol provides communications security over the Internet. The protocol allows client/server applications to communicate in a way that is designed to prevent eavesdropping, tampering, or message forgery. [STANDARDS-TRACK] 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 specification describes an optimized expression of the syntax of the Hypertext Transfer Protocol (HTTP). 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 messages on the same connection. It also introduces unsolicited push of representations from servers to clients. This document is an alternative to, but does not obsolete, the HTTP/1.1 message syntax. HTTP's existing semantics remain unchanged.

Thanks to Eliot Lear, Stephen Farrell, Guy Podjarny, Stephen Ludin, Erik Nygren, Paul Hoffman, Adam Langley, Will Chan and Richard Barnes for their feedback and suggestions. The Alt-Svc header field was influenced by the design of the Alternative-Protocol header in SPDY.