| Network Working Group | M. Nottingham |
| Internet-Draft | E. Wilde |
| Obsoletes: 7807 (if approved) | S. Dalal |
| Intended status: Standards Track | April 14, 2021 |
| Expires: October 16, 2021 |
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 define new error response formats for HTTP APIs.¶
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.¶
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.¶
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as “work in progress”.¶
This Internet-Draft will expire on October 16, 2021.¶
Copyright (c) 2021 IETF Trust and the persons identified as the document authors. All rights reserved.¶
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.¶
HTTP [RFC7230] status codes are sometimes not sufficient to convey enough information about an error to be helpful. While humans behind Web browsers can be informed about the nature of the problem with an HTML [HTML5] response body, non-human consumers of so-called "HTTP APIs" are usually not.¶
This specification defines simple JSON [RFC7159] and XML [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).¶
For example, 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; for example, triggering a transfer of more credit into the account.¶
This specification does this by identifying a specific type of problem (e.g., "out of credit") with a URI [RFC3986]; HTTP APIs can do this by nominating new URIs under their control, or by reusing existing ones.¶
Additionally, problem details 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 can be useful for support or forensic purposes.¶
The data model for problem details is a JSON [RFC7159] object; when formatted as a JSON document, it uses the "application/problem+json" media type. Appendix A defines how to express them in an equivalent XML format, which uses the "application/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. Likewise, in many situations, there is an appropriate HTTP status code that does not require extra detail to be conveyed.¶
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 redefine the semantics of existing HTTP status codes. Even if an application chooses not to use it to convey errors, reviewing its design can help guide the design decisions faced when conveying errors in an existing format.¶
When serialized as a JSON document, that format is identified with the "application/problem+json" media type.¶
For example, an HTTP response carrying JSON problem details:¶
HTTP/1.1 403 Forbidden
Content-Type: application/problem+json
Content-Language: en
{
"type": "https://example.com/probs/out-of-credit",
"title": "You do not have enough credit.",
"detail": "Your current balance is 30, but that costs 50.",
"instance": "/account/12345/msgs/abc",
"balance": 30,
"accounts": ["/account/12345",
"/account/67890"]
}
Here, the out-of-credit problem (identified by its type URI) indicates the reason for the 403 in "title", gives a reference for the specific problem occurrence with "instance", gives occurrence-specific details in "detail", and adds two extensions; "balance" conveys the account's balance, and "accounts" gives links where the account can be topped up.¶
The ability to convey problem-specific extensions allows more than one problem to be conveyed. For example:¶
HTTP/1.1 400 Bad Request
Content-Type: application/problem+json
Content-Language: en
{
"type": "https://example.net/validation-error",
"title": "Your request parameters didn't validate.",
"invalid-params": [ {
"name": "age",
"reason": "must be a positive integer"
},
{
"name": "color",
"reason": "must be 'green', 'red' or 'blue'"}
]
}
Note that this requires each of the subproblems to be similar enough to use the same HTTP status code. If they do not, the 207 (Multi-Status) [RFC4918] code could be used to encapsulate multiple status messages.¶
A problem details object can have the following members:¶
Consumers MUST use the "type" 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 do not have the ability to discover them (e.g., offline log analysis). Consumers SHOULD NOT automatically dereference the type URI.¶
The "status" 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 Section 5 for further caveats regarding its use.¶
Consumers can use the status member to determine what the original status code used by the generator was, in cases where it has been changed (e.g., by an intermediary or cache), and when message bodies persist without HTTP information. Generic HTTP software will still use the HTTP status code.¶
The "detail" member, if present, ought to focus on helping the client correct the problem, rather than giving debugging information.¶
Consumers SHOULD NOT parse the "detail" member for information; extensions are more suitable and less error-prone ways to obtain such information.¶
Problem type definitions MAY extend the problem details object with additional members.¶
For example, our "out of credit" problem above defines two such extensions -- "balance" and "accounts" to convey additional, problem-specific information.¶
Clients consuming problem details MUST ignore any such extensions that they don't recognize; this allows problem types to evolve and include additional information in the future.¶
Note that because extensions are effectively put into a namespace by the problem type, it is not possible to define new "standard" members without defining a new media type.¶
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. Designers of new problem types need to carefully consider the Security Considerations (Section 5), 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 in response to a PUT request is self-explanatory.¶
Finally, an application might 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; see [RFC7231], Section 5.3.2).¶
New problem type definitions MUST document:¶
Problem type definitions MAY specify the use of the Retry-After response header ([RFC7231], Section 7.1.3) in appropriate circumstances.¶
A problem's type URI SHOULD resolve to HTML [HTML5] 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 [RFC5988] 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 [RFC5234], Appendix B.1) and SHOULD consist of characters from ALPHA, DIGIT ([RFC5234], Appendix B.1), and "_" (so that it can be serialized in formats other than JSON), and they SHOULD be three characters or longer.¶
For example, if you 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 format.¶
To do so, you might look for an already-defined type URI that suits your purposes. If one is available, you can reuse that URI.¶
If one isn't available, you could mint and document a new type 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.¶
In summary: an instance URI will always identify a specific occurrence of a problem. On the other hand, type 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.¶
This specification reserves the use of one URI as a problem type:¶
The "about:blank" URI [RFC6694], when used as a problem type, indicates that the problem has no additional semantics beyond that of the HTTP status code.¶
When "about:blank" is used, the title SHOULD be the same as the recommended HTTP status phrase for that code (e.g., "Not Found" for 404, and so on), although it MAY be localized to suit client preferences (expressed with the Accept-Language request header).¶
Please note that according to how the "type" member is defined (Section 3.1), the "about:blank" URI is the default value for that member. Consequently, any problem details object not carrying an explicit "type" member implicitly uses this URI.¶
When defining a new problem type, the information included must be carefully vetted. Likewise, when actually generating a problem -- however it is serialized -- the details given must also be scrutinized.¶
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 "status" 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 (e.g., by a proxy or cache).¶
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, and virus scanners) are unlikely to know of or respect the status code conveyed in this member.¶
Some HTTP-based APIs use XML [XML] as their primary format convention. Such APIs can express problem details using the format defined in this appendix.¶
The RELAX NG schema [ISO-19757-2] for the XML format is as follows. Keep in mind that this schema is only meant as documentation, and not as a normative schema that captures all constraints of the XML format. Also, it would be possible to use other XML schema languages to define a similar set of constraints (depending on the features of the chosen schema language).¶
default namespace ns = "urn:ietf:rfc:7807"
start = problem
problem =
element problem {
( element type { xsd:anyURI }?
& element title { xsd:string }?
& element detail { xsd:string }?
& element status { xsd:positiveInteger }?
& element instance { xsd:anyURI }? ),
anyNsElement
}
anyNsElement =
( element ns:* { anyNsElement | text }
| attribute * { text })*
The media type for this format is "application/problem+xml".¶
Extension arrays and objects are serialized 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, the example above appears in XML as follows:¶
HTTP/1.1 403 Forbidden
Content-Type: application/problem+xml
Content-Language: en
<?xml version="1.0" encoding="UTF-8"?>
<problem xmlns="urn:ietf:rfc:7807">
<type>https://example.com/probs/out-of-credit</type>
<title>You do not have enough credit.</title>
<detail>Your current balance is 30, but that costs 50.</detail>
<instance>https://example.net/account/12345/msgs/abc</instance>
<balance>30</balance>
<accounts>
<i>https://example.net/account/12345</i>
<i>https://example.net/account/67890</i>
</accounts>
</problem>
Note that this format uses an XML namespace. This is primarily to allow embedding it into other XML-based formats; it does not imply that it can or should be extended with elements or attributes in 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 serialized into XML markup using only that namespace.¶
When using the XML format, it is possible to embed an XML processing instruction in the XML that instructs clients to transform the XML, using the referenced XSLT code [XSLT]. If this code is transforming the XML into (X)HTML, then it is possible to serve the XML format, and yet have clients capable of performing the transformation display human-friendly (X)HTML that is rendered and displayed at the client. Note that when using this method, it is advisable to use XSLT 1.0 in order to maximize the number of clients capable of executing the XSLT code.¶
In some situations, it can be advantageous to embed problem details in formats other than those described here. For example, an API that uses HTML [HTML5] might want to also use HTML for expressing its problem details.¶
Problem details can be embedded in other formats either by encapsulating one of the existing serializations (JSON or XML) into that format or by translating the model of a problem detail (as specified in Section 3) into the format's conventions.¶
For example, in HTML, a problem could be embedded by encapsulating JSON in a script tag:¶
<script type="application/problem+json">
{
"type": "https://example.com/probs/out-of-credit",
"title": "You do not have enough credit.",
"detail": "Your current balance is 30, but that costs 50.",
"instance": "/account/12345/msgs/abc",
"balance": 30,
"accounts": ["/account/12345",
"/account/67890"]
}
</script>
This specification does not make specific recommendations regarding embedding problem details in other formats; the appropriate way to embed them depends both upon the format in use and application of that format.¶
The authors would like to thank Jan Algermissen, Subbu Allamaraju, Mike Amundsen, Roy Fielding, Eran Hammer, Sam Johnston, Mike McCall, Julian Reschke, and James Snell for review of this specification.¶