title: On the use of HTTP as a Substrate docname: draft-ietf-httpbis-bcp56bis-00 date: {DATE} category: bcp obsoletes: 3205
ipr: trust200902 area: Applications and Real-Time workgroup: HTTP keyword: Internet-Draft keyword: HTTP keyword: Web keyword: substrate
stand_alone: yes pi: [toc, tocindent, sortrefs, symrefs, strict, compact, comments, inline]
ins: M. Nottingham
name: Mark Nottingham
organization:
email: mnot@mnot.net
uri: https://www.mnot.net/
informative: HTML5: target: https://html.spec.whatwg.org title: HTML - Living Standard author: - org: WHATWG FETCH: target: https://fetch.spec.whatwg.org title: Fetch - Living Standard author: - org: WHATWG
--- abstract
HTTP is often used as a substrate for other application protocols. This document specifies best practices for these protocols' use of HTTP.
--- note_Note_to_Readers_
Discussion of this draft takes place on the HTTP working group mailing list (ietf-http-wg@w3.org), which is archived at https://lists.w3.org/Archives/Public/ietf-http-wg/.
Working Group information can be found at http://httpwg.github.io/; source code and issues list for this draft can be found at bcp56bis .
--- middle
HTTP {{!RFC7230}} is often used as a substrate for other application protocols. This is done for a variety of reasons, including:
- familiarity by implementers, specifiers, administrators, developers and users,
- availability of a variety of client, server and proxy implementations,
- ease of use,
- ubiquity of Web browsers,
- reuse of existing mechanisms like authentication and encryption,
- presence of HTTP servers and clients in target deployments, and
- its ability to traverse firewalls.
The Internet community has a long tradition of protocol reuse, dating back to the use of Telnet {{?RFC0854}} as a substrate for FTP {{?RFC0959}} and SMTP {{?RFC2821}}. However, layering new protocols over HTTP brings its own set of issues:
- Should an application using HTTP define a new URL scheme? Use new ports?
- Should it use standard HTTP methods and status codes, or define new ones?
- How can the maximum value be extracted from the use of HTTP?
- How does it coexist with other uses of HTTP -- especially Web browsing?
- How can interoperability problems and "protocol dead ends" be avoided?
This document contains best current practices regarding the use of HTTP by applications other than Web browsing. {{used}} defines what applications it applies to; {{overview}} surveys the properties of HTTP that are important to preserve, and {{bp}} conveys best practices for those applications that do use HTTP.
It is written primarily to guide IETF efforts to define application protocols using HTTP for deployment on the Internet, but might be applicable in other situations. Note that the requirements herein do not necessarily apply to the development of generic HTTP extensions.
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 {{!RFC2119}} {{!RFC8174}} when, and only when, they appear in all capitals, as shown here.
Different applications have different goals when using HTTP. In this document, we say an application is using HTTP when any of the following conditions are true:
- The transport port in use is 80 or 443,
- The URL scheme "http" or "https" is used,
- The ALPN protocol ID {{!RFC7301}} "http/1.1", "h2" or "h2c" is used, or
- The message formats described in {{RFC7230}} and/or {{RFC7540}} are used in conjunction with the IANA registries defined for HTTP.
When an application is using HTTP, all of the requirements of the HTTP protocol suite (including but not limited to {{!RFC7230}}, {{!RFC7231}}, {{!RFC7232}}, {{!RFC7233}}, {{!RFC7234}}, {{!RFC7235}} and {{!RFC7540}}) are in force.
An application might not be using HTTP according to this definition, but still relying upon the HTTP specifications in some manner. For example, an application might wish to avoid re-specifying parts of the message format, but change others; or, it might want to use a different set of methods.
Such applications are referred to as protocols based upon HTTP in this document. These have more freedom to modify protocol operation, but are also likely to lose at least a portion of the benefits outlined above, as most HTTP implementations won't be easily adaptable to these changes, and as the protocol diverges from HTTP, the benefit of mindshare will be lost.
Protocols that are based upon HTTP MUST NOT reuse HTTP's URL schemes, transport ports, ALPN protocol IDs or IANA registries; rather, they are encouraged to establish their own.
There are many ways that HTTP applications are defined and deployed, and sometimes they are brought to the IETF for standardisation. In that process, what might be workable for deployment in a limited fashion isn't appropriate for standardisation and the corresponding broader deployment.
This section examines the facets of the protocol that are important to preserve in these situations.
When writing an application's specification, it's often tempting to specify exactly how HTTP is to be implemented, supported and used.
However, this can easily lead to an unintended profile of HTTP's behaviour. For example, it's common to see specifications with language like this:
A `200 OK` response means that the widget has successfully been updated.
This sort of specification is bad practice, because it is adding new semantics to HTTP's status codes and methods, respectively; a recipient -- whether it's an origin server, client library, intermediary or cache -- now has to know these extra semantics to understand the message.
Some applications even require specific behaviours, such as:
A `POST` request MUST result in a `201 Created` response.
This forms an expectation in the client that the response will always be 201 Created, when in
fact there are a number of reasons why the status code might differ in a real deployment. If the
client does not anticipate this, the application's deployment is brittle.
Much of the value of HTTP is in its generic semantics -- that is, the protocol elements defined by HTTP are potentially applicable to every resource, not specific to a particular context. Application-specific semantics are expressed in the payload; mostly, in the body, but also in header fields.
This allows a HTTP message to be examined by generic HTTP software (e.g., HTTP servers, intermediaries, client implementatiions), and its handling to be correctly determined. It also allows people to leverage their knowledge of HTTP semantics without special-casing them for a particular application.
Therefore, applications that use HTTP MUST NOT re-define, refine or overlay the semantics of defined protocol elements. Instead, they SHOULD focus their specifications on protocol elements that are specific to them; namely their HTTP resources.
See {{resource}} for details.
Another common practice is assuming that the HTTP server's name space (or a portion thereof) is exclusively for the use of a single application. This effectively overlays special, application-specific semantics onto that space, precludes other applications from using it.
As explained in {{!RFC7320}}, such "squatting" on a part of the URL space by a standard usurps the server's authority over its own resources, can cause deployment issues, and is therefore bad practice in standards.
Instead of statically defining URL paths, it is RECOMMENDED that applications using HTTP define links in payloads, to allow flexibility in deployment.
Using runtime links in this fashion has a number of other benefits. For example, navigating with a link allows a request to be routed to a different server without the overhead of a redirection, thereby supporting deployment across machines well. It becomes possible to "mix" different applications on the same server, and offers a natural path for extensibility, versioning and capability management.
The simplest possible use of HTTP is to POST data to a single URL, thereby effectively tunnelling through the protocol.
This "RPC" style of communication does get some benefit from using HTTP -- namely, message framing and the availability of implementations -- but fails to realise many others:
- Caching for server scalability, latency and bandwidth reduction, and reliability;
- Authentication and access control;
- Automatic redirection;
- Partial content to selectively request part of a response;
- Natural support for extensions and versioning through protocol extension; and
- The ability to interact with the application easily using a Web browser.
Using such a high-level protocol to tunnel simple semantics has downsides too; because of its more advanced capabilities, breadth of deployment and age, HTTP's complexity can cause interoperability problems that could be avoided by using a simpler substrate (e.g., WebSockets {{?RFC6455}}, if browser support is necessary, or TCP {{?RFC0793}} if not), or making the application be based upon HTTP, instead of using it (as defined in {{used}}).
Applications that use HTTP are encouraged to accommodate the various features that the protocol offers, so that their users receive the maximum benefit from it. This document does not require specific features to be used, since the appropriate design tradeoffs are highly specific to a given situation. However, following the practices in {{bp}} will help make them available.
This section contains best practices regarding the use of HTTP by applications, including practices for specific HTTP protocol elements.
When specifying the use of HTTP, an application SHOULD use {{!RFC7230}} as the primary reference; it is not necessary to reference all of the specifications in the HTTP suite unless there are specific reasons to do so (e.g., a particular feature is called out).
Applications using HTTP MAY specify a minimum version to be supported (HTTP/1.1 is suggested), and MUST NOT specify a maximum version.
Likewise, applications need not specify what HTTP mechanisms -- such as redirection, caching, authentication, proxy authentication, and so on -- are to be supported. Full featured support for HTTP SHOULD be taken for granted in servers and clients, and the application's function SHOULD degrade gracefully if they are not (although this might be achieved by informing the user that their task cannot be completed).
For example, an application can specify that it uses HTTP like this:
Foo Application uses HTTP {{!RFC7230}}. Implementations MUST support
HTTP/1.1, and MAY support later versions. Support for common HTTP
mechanisms such as redirection and caching are assumed.
HTTP Applications SHOULD focus on defining the following application-specific protocol elements:
- Media types {{!RFC6838}}, often based upon a format convention such as JSON {{?RFC7159}},
- HTTP header fields, as per {{headers}}, and
- The behaviour of resources, as identified by link relations {{!RFC5988}}.
By composing these protocol elements, an application can define a set of resources, identified by link relations, that implement specified behaviours, including:
- Retrieval of their state using GET, in one or more formats identified by media type;
- Resource creation or update using POST or PUT, with an appropriately identified request body format;
- Data processing using POST and identified request and response body format(s); and
- Resource deletion using DELETE.
For example, an application might specify:
Resources linked to with the "example-widget" link relation type are
Widgets. The state of a Widget can be fetched in the
"application/example-widget+json" format, and can be updated by PUT
to the same link. Widget resources can be deleted.
The "Example-Count" response header field on Widget representations
indicates how many Widgets are held by the sender.
The "application/example-widget+json" format is a JSON {{?RFC7159}}
format representing the state of a Widget. It contains links to
related information in the link indicated by the Link header field
value with the "example-other-info" link relation type.
In HTTP, URLs are opaque identifiers under the control of the server. As outlined in {{!RFC7320}}, standards cannot usurp this space, since it might conflict with existing resources, and constrain implementation and deployment.
In other words, applications that use HTTP MUST NOT associate application semantics with specific URL paths on arbitrary servers. Doing so inappropriately conflates the identity of the resource (its URL) with the capabilities that resource supports, bringing about many of the same interoperability problems that {{?RFC4367}} warns of.
For example, specifying that a "GET to the URL /foo retrieves a bar document" is bad practice. Likewise, specifying "The widget API is at the path /bar" violates {{!RFC7320}}.
Instead, applications that use HTTP are encouraged to ensure that URLs are discovered at runtime, allowing HTTP-based services to describe their own capabilities. One way to do this is to use typed links {{?RFC5988}} to convey the URIs that are in use, as well as the semantics of the resources that they identify. See {{resource}} for details.
Generally, a client will begin interacting with a given application server by requesting an initial document that contains information about that particular deployment, potentially including links to other relevant resources.
Applications that use HTTP SHOULD allow an arbitrary URL to be used as that entry point. For example, rather than specifying "the initial document is at "/foo/v1", they should allow a deployment to use any URL as the entry point for the application.
In cases where doing so is impractical (e.g., it is not possible to convey a whole URL, but only a hostname) applications that use HTTP MAY define a well-known URL {{?RFC5785}} as an entry point.
Applications that use HTTP will typically use the "http" and/or "https" URL schemes. "https" is preferred to provide authentication, integrity and confidentiality, as well as mitigate pervasive monitoring attacks {{?RFC7258}}.
However, application-specific schemes can be defined as well.
When defining an URL scheme for an application using HTTP, there are a number of tradeoffs and caveats to keep in mind:
-
Unmodified Web browsers will not support the new scheme. While it is possible to register new URL schemes with Web browsers (e.g. registerProtocolHandler() in {{HTML5}} Section 8.7.1.3, as well as several proprietary approaches), support for these mechanisms is not shared by all browsers, and their capabilities can vary.
-
Likewise, existing non-browser clients, intermediaries, servers and associated software will not recognise the new scheme, and might fail to operate. For example, a client library might fail to dispatch the request; a cache might refuse to store the response, and a proxy might fail to forward the request.
-
Because URLs occur in and are generated in HTTP artefacts commonly, often without human intervention (e.g., in the
Locationheader), it can be difficult to assure that the new scheme is used consistently. -
The resources identified by the new scheme will still be available with "http" and/or "https" URLs to clients. While it is possible to define the relationship between these resources in the new scheme's specification, existing HTTP software (such as clients, caches, intermediaries and servers) will not be available, so there is a danger of confusion when the "wrong" URL is used.
-
Features that rely upon the URL's origin {{?RFC6454}}, such as the Web's same-origin policy, will be impacted by a change of scheme.
-
HTTP-specific features such as cookies {{?RFC6265}}, authentication {{?RFC7235}}, caching {{?RFC7234}}, and CORS {{FETCH}} might or might not work correctly, depending on how they are defined and implemented. Generally, they are designed and implemented with an assumption that the URL will always be "http" or "https".
-
Web features that require a secure context {{?W3C.CR-secure-contexts-20160915}} will likely treat a new scheme as insecure.
See {{?RFC7595}} for more information about minting new URL schemes.
Applications that use HTTP SHOULD use the default port for the URL scheme in use. If it is felt that networks might need to distinguish the application's traffic for operational reasons, it MAY register a separate port, but be aware that this has privacy implications for that protocol's users. The impact of doing so MUST be documented in Security Considerations.
See {{?RFC7605}} for further guidance.
Applications that use HTTP MUST confine themselves to using registered HTTP methods such as GET, POST, PUT, DELETE, and PATCH.
New HTTP methods are rare; they are required to be registered with IETF Review (see {{!RFC7232}}), and are also required to be generic. That means that they need to be potentially applicable to all resources, not just those of one application.
While historically some applications (e.g., {{?RFC4791}}) has defined non-generic methods, {{!RFC7231}} now forbids this.
When it is believed that a new method is required, authors are encouraged to engage with the HTTP community early, and document their proposal as a separate HTTP extension, rather than as part of an application's specification.
Applications that use HTTP MUST only use registered HTTP status codes.
As with methods, new HTTP status codes are rare, and required (by {{!RFC7231}}) to be registered with IETF review. Similarly, HTTP status codes are generic; they are required (by {{!RFC7231}}) to be potentially applicable to all resources, not just to those of one application.
When it is believed that a new status code is required, authors are encouraged to engage with the HTTP community early, and document their proposal as a separate HTTP extension, rather than as part of an application's specification.
Status codes' primary function is to convey HTTP semantics for the benefit of generic HTTP software, not application-specific semantics. Therefore, applications MUST NOT specify additional semantics or refine existing semantics for status codes.
In particular, specifying that a particular status code has a specific meaning in the context of an application is harmful, as these are not generic semantics, since the consumer needs to be in the context of the application to understand them.
Furthermore, applications using HTTP MUST NOT re-specify the semantics of HTTP status codes, even if it is only by copying their definition. They MUST NOT require specific reason phrases to be used; the reason phrase has no function in HTTP, and is not guaranteed to be preserved by implementations. The reason phrase is not carried in the {{RFC7540}} message format.
Typically, applications using HTTP will convey application-specific information in the message body and/or HTTP header fields, not the status code.
Specifications sometimes also create a "laundry list" of potential status codes, in an effort to be
helpful. The problem with doing so is that such a list is never complete; for example, if a network
proxy is interposed, the client might encounter a 407 Proxy Authentication Required response; or,
if the server is rate limiting the client, it might receive a 429 Too Many Requests response.
Since the list of HTTP status codes can be added to, it's safer to refer to it directly, and point
out that clients SHOULD be able to handle all applicable protocol elements gracefully (i.e.,
falling back to the generic n00 semantics of a given status code; e.g., 499 can be safely
handled as 400 by clients that don't recognise it).
Applications that use HTTP MAY define new HTTP header fields, following the advice in {{!RFC7231}}, Section 8.3.1.
Typically, using HTTP header fields is appropriate in a few different situations:
- Their content is useful to intermediaries (who often wish to avoid parsing the body), and/or
- Their content is useful to generic HTTP software (e.g., clients, servers), and/or
- It is not possible to include their content in the message body (usually because a format does not allow it).
If none of these motivations apply, using a header field is NOT RECOMMENDED.
New header fields MUST be registered, as per {{!RFC7231}} and {{!RFC3864}}.
It is RECOMMENDED that header field names be short (even when HTTP/2 header compression is in effect, there is an overhead) but appropriately specific. In particular, if a header field is specific to an application, an identifier for that application SHOULD form a prefix to the header field name, separated by a "-".
The semantics of existing HTTP header fields MUST NOT be re-defined without updating their
registration or defining an extension to them (if allowed). For example, an application using HTTP
cannot specify that the Location header has a special meaning in a certain context.
See {{state}} for requirements regarding header fields that carry application state (e.g,. Cookie).
Modern Web browsers constrain the ability of content from one origin (as defined by {{?RFC6454}}) to access resources from another, to avoid the "confused deputy" problem. As a result, applications that wish to expose cross-origin data to browsers will need to implement {{!W3C.REC-cors-20140116}}.
Applications that use HTTP MAY use stateful cookies {{?RFC6265}} to identify a client and/or store client-specific data to contextualise requests.
If it is only necessary to identify clients, applications that use HTTP MAY use HTTP authentication {{?RFC7235}}; if the Basic authentication scheme {{?RFC7617}} is used, it MUST NOT be used with the 'http' URL scheme.
In either case, it is important to carefully specify the scoping and use of these mechanisms; if they expose sensitive data or capabilities (e.g., by acting as an ambient authority), exploits are possible. Mitigations include using a request-specific token to assure the intent of the client.
This document has no requirements for IANA.
{{state}} discusses the impact of using stateful mechanisms in the protocol as ambient authority, and suggests a mitigation.
{{scheme}} requires support for 'https' URLs, and discourages the use of 'http' URLs, to provide authentication, integrity and confidentiality, as well as mitigate pervasive monitoring attacks.
Applications that use HTTP in a manner that involves modification of implementations -- for example, requiring support for a new URL scheme, or a non-standard method -- risk having those implementations "fork" from their parent HTTP implementations, with the possible result that they do not benefit from patches and other security improvements incorporated upstream.
--- back
RFC3205 captured the Best Current Practice in the early 2000's, based on the concerns facing protocol designers at the time. Use of HTTP has changed considerably since then, and as a result this document is substantially different. As a result, the changes are too numerous to list individually.