]> Akamai

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

EMC

erik.wilde@emc.com

status HTTP error problem API JSON XML This document defines a "problem detail" as a way to carry machine-readable details of errors in a HTTP response, to avoid the need to invent new error response formats for HTTP APIs. This draft should be discussed on the apps-discuss mailing list.

HTTP status codes are sometimes not sufficient to convey helpful information about an error. While humans behind Web browsers can be informed about the nature of the problem with an HTML response body, non-human consumers of so-called "HTTP APIs" are usually not. This specification defines simple JSON and XML document formats to suit this purpose. They are designed to be reused by HTTP APIs, which can identify distinct "problem types" specific to their needs. Thus, API clients can be informed of both the high-level error class (using the status code) and the finer-grained details of the problem (using one of these formats). Consider a response that indicates that the client's account doesn't have enough credit. The 403 Forbidden status code might be deemed most appropriate to use, as it will inform HTTP-generic software (such as client libraries, caches and proxies) of the general semantics of the response. However, that doesn't give the API client enough information about why the request was forbidden, the applicable account balance, or how to correct the problem. If these details are included in the response body in a machine-readable format, the client can treat it appropriately. This specification does this by identifying a specific type of problem (e.g., "out of credit") with a URI ; APIs can do this by nominating new URIs under their control, or by reusing existing ones. Additionally, problems can contain other information, such as a URI that identifies the specific occurrence of the problem (effectively giving an identifier to the concept "The time Joe didn't have enough credit last Thursday"), which may be useful for support or forensic purposes. See below for a full list. The data model for problem details is a JSON object; when formatted as a JSON document, it uses the "application/api-problem+json" media type. defines how to express them in an equivalent XML format, which uses the "application/api-problem+xml" media type. Note that problem details are (naturally) not the only way to convey the details of a problem in HTTP; if the response is still a representation of a resource, for example, it's often preferable to accommodate describing the relevant details in that application's format. Instead, the aim of this specification is to define common error formats for those applications that need one, so that they aren't required to define their own, or worse, tempted to re-define the semantics of existing HTTP status codes.

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 .

The canonical model for problem details is a JSON object. When serialised as a JSON document, that format is identified with the "application/api-problem+json" media type. For example, a HTTP response carrying JSON problem details:

Here, the out-of-credit problem (identified by its problemType URI) indicates the reason for the 403 in "title", gives a reference for the specific problem occurrence with "problemInstance", gives occurrence-specific details in "detail", and adds two extensions; "balance" conveys the account's balance, and "account" gives a link where the account can be topped up. Note that "problemType" is case-sensitive in the JSON object, as are all other member names.

A problem details object MUST have the following members: "problemType" (string) - An absolute URI that identifies the problem type. When dereferenced, it SHOULD provide human-readable documentation for the problem type (e.g., using HTML). "title" (string) - A short, human-readable summary of the problem type. It SHOULD NOT change from occurrence to occurrence of the problem, except for purposes of localisation. Consumers MUST use the problemType string as the primary identifier for the problem type; the title string is advisory, and included only for users who are not aware of the semantics of the URI, and don't have the ability to discover them (e.g., offline log analysis). Consumers SHOULD NOT automatically dereference the problemType URI.

Furthermore, a problem details object MAY have the following members: "httpStatus" (number) - The HTTP status code set by the origin server for this occurrence of the problem. "detail" (string) - An human readable explanation specific to this occurrence of the problem. "problemInstance" (string) - An absolute URI that identifies the specific occurrence of the problem. It may or may not yield further information if dereferenced. The httpStatus member, if present, is only advisory; it conveys the HTTP status code used for the convenience of the consumer. Generators MUST use the same status code in the actual HTTP response, to assure that generic HTTP software that does not understand this format still behaves correctly. See for further caveats regarding its use. The detail member, if present, SHOULD focus on helping the client correct the problem, rather than giving debugging information. Consumers SHOULD NOT be parse the detail member for information; extensions are more suitable and less error-prone ways to obtain such information.

Finally, problem type definitions MAY extend the problem details object with additional members. Clients consuming problem details MUST ignore any such extensions that they don't recognise; this allows problem types to evolve and include additional information in the future.

When an HTTP API needs to define a response that indicates an error condition, it might be appropriate to do so by defining a new problem type. Before doing so, it's important to understand what they are good for, and what's better left to other mechanisms. Problem details are not a debugging tool for the underlying implementation; rather, they are a way to expose greater detail about the HTTP interface itself. New problem types need to carefully consider the Security Considerations (); in particular the risk of exposing attack vectors by exposing implementation internals through error messages. Likewise, truly generic problems – i.e., conditions that could potentially apply to any resource on the Web – are usually better expressed as plain status codes. For example, a "write access disallowed" problem is probably unnecessary, since a 403 Forbidden status code on a PUT request is self-explanatory. Finally, an application may have a more appropriate way to carry an error in a format that it already defines. Problem details are intended to avoid the necessity of establishing new "fault" or "error" document formats, not to replace existing domain-specific formats. That said, it is possible to add support for problem details to existing HTTP APIs using HTTP content negotiation (e.g., using the Accept request header to indicate a preference for this format). New problem type definitions MUST document: A problemType URI (typically, with the "http" scheme), A title that appropriately describes it (think short), and The HTTP status code for it to be used with. Problem types MAY specify the use of the Retry-After response header in appropriate circumstances. A problem's problemType URI SHOULD resolve to HTML documentation that explains how to resolve the problem. A problem type definition MAY specify additional members on the Problem Details object. For example, an extension might use typed links to another resource that can be used by machines to resolve the problem. If such additional members are defined, their names SHOULD start with a letter (ALPHA, as per and SHOULD consist of characters from ALPHA, DIGIT and "_" (so that it can be serialised in formats other than JSON), and SHOULD be three characters or longer.

For example, if your are publishing an HTTP API to your online shopping cart, you might need to indicate that the user is out of credit (our example from above), and therefore cannot make the purchase. If you already have an application-specific format that can accommodate this information, it's probably best to do that. However, if you don't, you might consider using one of the problem details formats; JSON if your API is JSON-based, or XML if it uses that convention. To do so, you might look for an already-defined problemType URI that suits your purposes. If one is available, you can reuse that URI by documenting use of both this specification and that specific URI value in your API. If one isn't available, you'd reference this specification, mint and document a new problemType URI (which ought to be under your control and stable over time), an appropriate title and the HTTP status code that it will be used with, along with what it means and how it should be handled. This information could appear within the rest of your API's documentation. In summary: a problemInstance URI will always identify a specific occurrence of a problem. On the other hand, problemType URIs can be reused if an appropriate description of a problem type is already available someplace else, or they can be created for new problem types.

When defining a new problem type, the information included must be carefully vetted. Likewise, when actually generating a problem – however it is serialised – the details given must also be scrutinised. Risks include leaking information that can be exploited to compromise the system, access to the system, or the privacy of users of the system. Generators providing links to occurrence information are encouraged to avoid making implementation details such as a stack dump available through the HTTP interface, since this can expose sensitive details of the server implementation, its data, and so on. The "httpStatus" member duplicates the information available in the HTTP status code itself, thereby bringing the possibility of disagreement between the two. Their relative precedence is not clear, since a disagreement might indicate that (for example) an intermediary has modified the HTTP status code in transit. As such, those defining problem types as well as generators and consumers of problems need to be aware that generic software (such as proxies, load balancers, firewalls, virus scanners) are unlikely to know of or respect the status code conveyed in this member.

This specification defines two new Internet media types :

Intended usage: COMMON Restrictions on usage: None. Author: Mark Nottingham Change controller: IESG ]]>

Intended usage: COMMON Restrictions on usage: None. Author: Mark Nottingham Change controller: IESG ]]>

The authors would like to thank Jan Algermissen, Mike Amundsen, Subbu Allamaraju, Roy Fielding, Sam Johnston, Mike McCall, Julian Reschke, and James Snell for early review of this specification (even if some disagree with parts of it).

&rfc2119; &rfc2616; &rfc3986; &rfc4627; &rfc5234; &xml; International Organization for Standardization &rfc3023; &rfc5988; &rfc6838;

Some HTTP-based APIs use XML as their primary format convention. Such APIs MAY express problem details using the format defined in this appendix. The OPTIONAL RELAX NG schema for the XML format is:

&problem-schema;

The media type for this format is "application/api-problem+xml". Extension arrays and objects can be serialised into the XML format by considering an element containing a child or children to represent an object, except for elements that contain only child element(s) named 'i', which are considered arrays. For example, an alternate version of the example above would appear in XML as:

http://example.com/probs/out-of-credit Your current balance is 30, but that costs 50. http://example.net/account/12345/msgs/abc 30 http://example.net/account/12345 http://example.net/account/67890 ]]>

Note that this format uses an XML Namespace. This is primarily to allow embedding it into other formats; it does not imply that it can be extended with content from other namespaces. The RELAX NG schema explicitly only allows elements from the one namespace used in the XML format. Any extension arrays and objects MUST be serialised using that namespace.