API Guild Logo

Other formats: PDF, EPUB3

Table of Contents

1. Introduction

Zalando’s software architecture centers around decoupled microservices that provide functionality via RESTful APIs with a JSON payload. Small engineering teams own, deploy and operate these microservices in their AWS (team) accounts. Our APIs most purely express what our systems do, and are therefore highly valuable business assets. Designing high-quality, long-lasting APIs has become even more critical for us since we started developing our new open platform strategy, which transforms Zalando from an online shop into an expansive fashion platform. Our strategy emphasizes developing lots of public APIs for our external business partners to use via third-party applications.

With this in mind, we’ve adopted “API First” as one of our key engineering principles. Microservices development begins with API definition outside the code and ideally involves ample peer-review feedback to achieve high-quality APIs. API First encompasses a set of quality-related standards and fosters a peer review culture including a lightweight review procedure. We encourage our teams to follow them to ensure that our APIs:

  • are easy to understand and learn

  • are general and abstracted from specific implementation and use cases

  • are robust and easy to use

  • have a common look and feel

  • follow a consistent RESTful style and syntax

  • are consistent with other teams’ APIs and our global architecture

Ideally, all Zalando APIs will look like the same author created them.

Conventions Used in These Guidelines

The requirement level keywords "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" used in this document (case insensitive) are to be interpreted as described in RFC 2119.

Zalando specific information

The purpose of our “RESTful API guidelines” is to define standards to successfully establish “consistent API look and feel” quality. The API Guild [internal link] drafted and owns this document. Teams are responsible to fulfill these guidelines during API development and are encouraged to contribute to guideline evolution via pull requests.

These guidelines will, to some extent, remain work in progress as our work evolves, but teams can confidently follow and trust them.

In case guidelines are changing, following rules apply:

  • existing APIs don’t have to be changed, but we recommend it

  • clients of existing APIs have to cope with these APIs based on outdated rules

  • new APIs have to respect the current guidelines

Furthermore you should keep in mind that once an API becomes public externally available, it has to be re-reviewed and changed according to current guidelines - for sake of overall consistency.

2. Principles

API as a Product

As mentioned above, Zalando is transforming from an online shop into an expansive fashion platform comprising a rich set of products following a Software as a Platform (SaaP) model for our business partners. As a company we want to deliver products to our (internal and external) customers which can be consumed like a service.

Platform products provide their functionality via (public) APIs; hence, the design of our APIs should be based on the API as a Product principle:

  • Treat your API as product and understand the needs of its customers

  • Take ownership and advocate for the customer and continuous improvement

  • Emphasize easy understanding, discovery and usage of APIs; design APIs irresistible for client engineers

  • Actively improve and maintain API consistency over the long term

  • Make use of customer feedback and provide service level support

RESTful API as a Product makes the difference between enterprise integration business and agile, innovative product service business built on a platform of APIs.

Based on your concrete customer use cases, you should carefully check the trade-offs of API design variants and avoid short-term server side implementation optimizations at the expense of unnecessary client side obligations and have a high attention on API quality and client developer experience.

API as a Product is closely related to our API First principle (see next chapter) which is more focused on how we engineer high quality APIs.

API Design Principles

Comparing SOA web service interfacing style of SOAP vs. REST, the former tend to be centered around operations that are usually use-case specific and specialized. In contrast, REST is centered around business (data) entities exposed as resources that are identified via URIs and can be manipulated via standardized CRUD-like methods using different representations, self-descriptive messages and hypermedia. RESTful APIs tend to be less use-case specific and comes with less rigid client / server coupling and are more suitable as a platform interface being open for diverse client applications.

  • We prefer REST-based APIs with JSON payloads

  • We prefer systems to be truly RESTful [1]

We apply the RESTful web service principles to all kind of application components, whether they provide functionality via the Internet or via the intranet as larger application elements. We strive to build interoperating distributed systems that different teams can evolve in parallel.

An important principle for (RESTful) API design and usage is Postel’s Law, aka the Robustness Principle (RFC 1122):

  • Be liberal in what you accept, be conservative in what you send

Readings: Read the following to gain additional insight on the RESTful service architecture paradigm and general RESTful API design style:

3. General Guidelines

The titles are marked with the corresponding labels: Must:, Should:, May:.

Must: Follow API First Principle

As mentioned in the introduction, API First is one of our engineering and architecture principles. In a nutshell API First requires two aspects:

  • define APIs outside the code first using a standard specification language

  • get early review feedback from peers and client developers

By defining APIs outside the code, we want to facilitate early review feedback and also a development discipline that focus service interface design on…​

  • profound understanding of the domain and required functionality

  • generalized business entities / resources, i.e. avoidance of use case specific APIs

  • clear separation of WHAT vs. HOW concerns, i.e. abstraction from implementation aspects — APIs should be stable even if we replace complete service implementation including its underlying technology stack

Moreover, API definitions with standardized specification format also facilitate…​

  • single source of truth for the API specification; it is a crucial part of a contract between service provider and client users

  • infrastructure tooling for API discovery, API GUIs, API documents, automated quality checks

An element of API First are also this API Guidelines and a lightweight API review process [internal link] as to get early review feedback from peers and client developers. Peer review is important for us to get high quality APIs, to enable architectural and design alignment and to supported development of client applications decoupled from service provider engineering life cycle.

It is important to learn, that API First is not in conflict with the agile development principles that we love. Service applications should evolve incrementally — and so its APIs. Of course, our API specification will and should evolve iteratively in different cycles; however, each starting with draft status and early team and peer review feedback. API may change and profit from implementation concerns and automated testing feedback. API evolution during development life cycle may include breaking changes for not yet productive features and as long as we have aligned the changes with the clients. Hence, API First does not mean that you must have 100% domain and requirement understanding and can never produce code before you have defined the complete API and get it confirmed by peer review. On the other hand, API First obviously is in conflict with the bad practice of publishing API definition and asking for peer review after the service integration or even the service productive operation has started. It is crucial to request and get early feedback — as early as possible, but not before the API changes are comprehensive with focus to the next evolution step and have a certain quality (including API Guideline compliance), already confirmed via team internal reviews.

Must: Provide API Reference Definition using OpenAPI

We use the OpenAPI specification (aka Swagger spec) as standard for our REST API definitions. You have to provide the API definition in YAML format (instead of JSON) for the OpenAPI API definition files due to its improved readability.

Please stick to version 2.0 of the specification for now, until we succeed to update all our tooling to support the upcoming version 3.0.

We also call the OpenAPI API definition the "API Reference definition" (or "API definition"); as a reference manual it provides all information needed by an experienced API client developer to use this API.

The OpenAPI API specification file should be subject of version control together with source code management. Services also have to support an endpoint to access the API Reference definition for their external API(s).

Furthermore you should make use of Zally, our own API linter which supports you in designing a guideline compliant API. You find more information in the Github repository itself and in API review procedures [internal link].

Should: Provide User Manual Documentation

In addition to the API as OpenAPI Reference Definition, it’s good practice to provide an API User Manual documentation to improve client developer experience, especially of engineers that are less experienced in using this API. A helpful API User Manual typically describes the following API aspects:

  • API’s scope, purpose and use cases

  • concrete examples of API usage

  • edge cases, error situation details and repair hints

  • architecture context and major dependencies - including figures and sequence flows

The User Manual must be posted online, e.g. via GitHub Enterprise pages, on specific team web servers, or as a Google doc. And don’t forget to include a link to this user manual into your OpenAPI definition using the “externalDocs” property.

4. Security

Must: Secure Endpoints with OAuth 2.0

Every API endpoint needs to be secured using OAuth 2.0. Please refer to the official OpenAPI spec on how to specify security definitions in you API specification or take a look at the following example.

    type: oauth2
    flow: password
    tokenUrl: https://token.services.auth.zalando.com/oauth2/access_token
      fulfillment-order-service.read: Access right needed to read from the fulfillment order service.
      fulfillment-order-service.write: Access right needed to write to the fulfillment order service.

The example defines OAuth2 with password flow as security standard used for authentication when accessing endpoints; additionally, there are two API access rights defined via the scopes section for later endpoint authorization usage - please see next section.

It makes little sense specifying the flow to retrieve OAuth tokens in the securityDefinitions section, as API endpoints should not care, how OAuth tokens were created. Unfortunately the flow field is mandatory and cannot be ommited. API endpoints should always set flow: password and ignore this information.

Must: Define and Assign Access Rights (Scopes)

Every API needs to define access rights, called scopes here, and every endpoint needs to have at least one scope assigned. Scopes are defined by name and description per API specification, as shown in the previous section. Please refer to the following rules when creating scope names:

<api-scope> ::= <api-standard-scope> |            -- should be sufficient for majority of use cases
                <api-resource-specific-scope> |   -- for special security access differentiation use cases
                <api-pseudo-scope>                -- used to explicitly indicate that access is not restricted

<api-standard-scope>          ::= <application-id>.<access-type>
<api-resource-specific-scope> ::= <application-id>.<resource-id>.<access-type>
<api-pseudo-scope>            ::= uid

<application-id> ::= <as defined via STUPS>
<access-type>    ::= read | write           -- might be extended in future
<resource-id>    ::= <free identifier following application-id syntax>

APIs should stick to standard scopes by default — for the majority of use cases, restricting access to specific APIs (with read vs. write differentiation) is sufficient for controlling access for client types like merchant or retailer business partners, customers or operational staff. We want to avoid too many, fine grained scopes increasing governance complexity without real value add. In some situations, where the API serves different types of resources for different owners, resource specific scopes may make sense.

Some examples for standard and resource-specific scopes:

Application ID Resource ID Access Type Example















After scopes names are defined and the scope is declared in the security definition at the top of an API specification, it should be assigned to each API operation by specifying a security requirement like this:

      summary: Retrieves a sales order
        - oauth2:
          - sales-order-service.sales_order.read

In very rare cases a whole API or some selected endpoints may not require specific access control. However, to make this explicit you should assign the uid pseudo access right scope in this case. It is the user id and always available as OAuth2 default scope.

      summary: Provides public information about ...
               Accessible by any user; no access rights needed.
        - oauth2:
          - uid

Hint: you need not explicitly define the "Authorization" header; it is a standard header so to say implicitly defined via the security section.

5. Compatibility

Must: Don’t Break Backward Compatibility

Change APIs, but keep all consumers running. Consumers usually have independent release lifecycles, focus on stability, and avoid changes that do not provide additional value. APIs are contracts between service providers and service consumers that cannot be broken via unilateral decisions.

There are two techniques to change APIs without breaking them:

  • follow rules for compatible extensions

  • introduce new API versions and still support older versions

We strongly encourage using compatible API extensions and discourage versioning. The below guideline rules for service providers and consumers enable us (having Postel’s Law in mind) to make compatible changes without versioning.

Hint: Please note that the compatibility guarantees are for the "on the wire" format. Binary or source compatibility of code generated from an API definition is not covered by these rules. If client implementations update their generation process to a new version of the API definition, it has to be expected that code changes are necessary.

Should: Prefer Compatible Extensions

API designers should apply the following rules to evolve RESTful APIs for services in a backward-compatible way:

  • Add only optional, never mandatory fields.

  • Never change the semantic of fields (e.g. changing the semantic from customer-number to customer-id, as both are different unique customer keys)

  • Input fields may have (complex) constraints being validated via server-side business logic. Never change the validation logic to be more restrictive and make sure that all constraints a clearly defined in description.

  • Enum ranges can be reduced when used as input parameters, only if the server is ready to accept and handle old range values too. Enum range can be reduced when used as output parameters.

  • Enum ranges cannot not be extended when used for output parameters — clients may not be prepared to handle it. However, enum ranges can be extended when used for input parameters.

  • Use x-extensible-enum, if range is used for output parameters and likely to be extended with growing functionality. It defines an open list of explicit values and clients must be agnostic to new values.

  • Support redirection in case an URL has to change (301 Moved Permanently).

Must: Prepare Clients To Not Crash On Compatible API Extensions

Service clients should apply the robustness principle:

  • Be conservative with API requests and data passed as input, e.g. avoid to exploit definition deficits like passing megabytes for strings with unspecified maximum length.

  • Be tolerant in processing and reading data of API responses, more specifically…​

Service clients must be prepared for compatible API extensions of service providers:

  • Be tolerant with unknown fields in the payload (see also Fowler’s "TolerantReader" post), i.e. ignore new fields but do not eliminate them from payload if needed for subsequent PUT requests.

  • Be prepared that x-extensible-enum return parameter may deliver new values; either be agnostic or provide default behavior for unknown values.

  • Be prepared to handle HTTP status codes not explicitly specified in endpoint definitions. Note also, that status codes are extensible. Default handling is how you would treat the corresponding x00 code (see RFC7231 Section 6).

  • Follow the redirect when the server returns HTTP status 301 Moved Permanently.

Should: Design APIs Conservatively

Designers of service provider APIs should be conservative and accurate in what they accept from clients:

  • Unknown input fields in payload or URL should not be ignored; servers should provide error feedback to clients via an HTTP 400 response code.

  • Be accurate in defining input data constraints (like formats, ranges, lengths etc.) — and check constraints and return dedicated error information in case of violations.

  • Prefer being more specific and restrictive (if compliant to functional requirements), e.g. by defining length range of strings. It may simplify implementation while providing freedom for further evolution as compatible extensions.

Not ignoring unknown input fields is a specific deviation from Postel’s Law (e.g. see also
The Robustness Principle Reconsidered) and a strong recommendation. Servers might want to take different approach but should be aware of the following problems and be explicit in what is supported:

  • Ignoring unknown input fields is actually not an option for PUT, since it becomes asymmetric with subsequent GET response and HTTP is clear about the PUT "replace" semantics and default roundtrip expectations (see RFC7231 Section 4.3.4). Note, accepting (i.e. not ignoring) unknown input fields and returning it in subsequent GET responses is a different situation and compliant to PUT semantics.

  • Certain client errors cannot be recognized by servers, e.g. attribute name typing errors will be ignored without server error feedback. The server cannot differentiate between the client intentionally providing an additional field versus the client sending a mistakenly named field, when the client’s actual intent was to provide an optional input field.

  • Future extensions of the input data structure might be in conflict with already ignored fields and, hence, will not be compatible, i.e. break clients that already use this field but with different type.

In specific situations, where a (known) input field is not needed anymore, it either can stay in the API definition with "not used anymore" description or can be removed from the API definition as long as the server ignores this specific parameter.

Must: Always Return JSON Objects As Top-Level Data Structures To Support Extensibility

In a response body, you must always return a JSON object (and not e.g. an array) as a top level data structure to support future extensibility. JSON objects support compatible extension by additional attributes. This allows you to easily extend your response and e.g. add pagination later, without breaking backwards compatibility.

Must: Treat Open API Definitions As Open For Extension By Default

The Open API 2.0 specification is not very specific on default extensibility of objects, and redefines JSON-Schema keywords related to extensibility, like additionalProperties. Following our overall compatibility guidelines, Open API object definitions are considered open for extension by default as per Section 5.18 "additionalProperties" of JSON-Schema.

When it comes to Open API 2.0, this means an additionalProperties declaration is not required to make an object definition extensible:

  • API clients consuming data must not assume that objects are closed for extension in the absence of an additionalProperties declaration and must ignore fields sent by the server they cannot process. This allows API servers to evolve their data formats.

  • For API servers receiving unxpected data, the situation is slightly different. Instead of ignoring fields, servers may reject requests whose entities contain undefined fields in order to signal to clients that those fields would not be stored on behalf of the client. API designers must document clearly how unexpected fields are handled for PUT, POST and PATCH requests.

API formats must not declare additionalProperties to be false, as this prevents objects being extended in the future.

Note that this guideline concentrates on default extensibility and does not exclude the use of additionalProperties with a schema as a value, which might be appropriate in some circumstances.

Should: Used Open-Ended List of Values (x-extensible-enum) Instead of Enumerations

Enumerations are per definition closed sets of values, that are assumed to be complete and not intended for extension. This closed principle of enumerations imposes compatibility issues when an enumeration must be extended. To avoid these issues, we strongly recommend to use an open-ended list of values instead of an enumeration unless:

  1. the API has full control of the enumeration values, i.e. the list of values does not depend on any external tool or interface, and

  2. the list of value is complete with respect to any thinkable and unthinkable future feature.

To specify an open-ended list of values use the marker x-extensible-enum as follows:

  type: string
    - parcel
    - letter
    - email

Note: x-extensible-enum is not JSON Schema conform but will be ignored by most tools.

Should: Avoid Versioning

When changing your RESTful APIs, do so in a compatible way and avoid generating additional API versions. Multiple versions can significantly complicate understanding, testing, maintaining, evolving, operating and releasing our systems (supplementary reading).

If changing an API can’t be done in a compatible way, then proceed in one of these three ways:

  • create a new resource (variant) in addition to the old resource variant

  • create a new service endpoint — i.e. a new application with a new API (with a new domain name)

  • create a new API version supported in parallel with the old API by the same microservice

As we discourage versioning by all means because of the manifold disadvantages, we suggest to only use the first two approaches.

Must: Use Media Type Versioning

When API versioning is unavoidable, you have to design your multi-version RESTful APIs using media type versioning (instead of URI versioning, see below). Media type versioning is less tightly coupled since it supports content negotiation and hence reduces complexity of release management.

Media type versioning: Here, version information and media type are provided together via the HTTP Content-Type header — e.g. application/x.zalando.cart+json;version=2. For incompatible changes, a new media type version for the resource is created. To generate the new representation version, consumer and producer can do content negotiation using the HTTP Content-Type and Accept headers. Note: This versioning only applies to the request and response content schema, not to URI or method semantics.

In this example, a client wants only the new version of the response:

Accept: application/x.zalando.cart+json;version=2

A server responding to this, as well as a client sending a request with content should use the Content-Type header, declaring that one is sending the new version:

Content-Type: application/x.zalando.cart+json;version=2

Using header versioning should:

  • include versions in request and response headers to increase visibility

  • include Content-Type in the Vary header to enable proxy caches to differ between versions

Hint: OpenAPI currently doesn’t support content negotiation, though a comment in this issue mentions a workaround (using a fragment identifier that gets stripped off). Another way would be to document just the new version, but let the server accept the old one (with the previous content-type).

Until an incompatible change is necessary, it is recommended to stay with the standard application/json media type.

Must: Do Not Use URI Versioning

With URI versioning a (major) version number is included in the path, e.g. /v1/customers. The consumer has to wait until the provider has been released and deployed. If the consumer also supports hypermedia links — even in their APIs — to drive workflows (HATEOAS), this quickly becomes complex. So does coordinating version upgrades — especially with hyperlinked service dependencies — when using URL versioning. To avoid this tighter coupling and complexer release management we do not use URI versioning, and go instead with media type versioning and content negotiation (see above).

Should: Provide Version Information in OpenAPI Documentation

Only the documentation, not the API itself, needs version information.


"swagger": "2.0",
"info": {
  "title": "Parcel service API",
  "description": "API for <...>",
  "version": "1.0.0",

During a (possibly) long-running API review phase you need different versions of the API description. These versions may include changes that are incompatible with earlier draft versions. So we apply the following version schema MAJOR.MINOR.DRAFT that increments the…​

  • MAJOR version, when you make incompatible API changes

  • MINOR version, when you add functionality in a backwards-compatible manner

  • DRAFT version, when you make changes during the review phase that are not related to production releases

We recommend using the DRAFT version only for unreleased API definitions that are still under review; for example:

version 1.4.0  -- current version
version 1.4.1  -- first draft and call for review of API extensions compatible with 1.4.0
version 1.4.2  -- second draft and call for review of API extensions that are still compatible with 1.4.0 but possibly incompatible with 1.4.1
version 1.5.0  -- approved version for implementation and release
version 1.5.1  -- first draft for next review and API change cycle; compatible with 1.4.0 and 1.5.0

Hint: This versioning scheme differs in the less strict DRAFT aspect from semantic version information used for released APIs and service applications.

6. JSON Guidelines

These guidelines provides recommendations for defining JSON data at Zalando. JSON here refers to RFC 7159 (which updates RFC 4627), the “application/json” media type and custom JSON media types defined for APIs. The guidelines clarifies some specific cases to allow Zalando JSON data to have an idiomatic form across teams and services.

Must: Property names must be snake_case (and never camelCase)

No established industry standard exists, but many popular Internet companies prefer snake_case: e.g. GitHub, Stack Exchange, Twitter. Others, like Google and Amazon, use both - but not only camelCase. It’s essential to establish a consistent look and feel such that JSON looks as if it came from the same hand.

Must: Property names must match ^[a-z_][a-z_0-9]*$

Property names are restricted to ASCII strings. The first character must be a letter, or an underscore, and subsequent characters can be a letter, an underscore, or a number.

(It is recommended to use _ at the start of property names only for keywords like _links.)

Should: Array names should be pluralized

To indicate they contain multiple values prefer to pluralize array names. This implies that object names should in turn be singular.

Must: Boolean property values must not be null

Schema based JSON properties that are by design booleans must not be presented as nulls. A boolean is essentially a closed enumeration of two values, true and false. If the content has a meaningful null value, strongly prefer to replace the boolean with enumeration of named values or statuses - for example accepted_terms_and_conditions with true or false can be replaced with terms_and_conditions with values yes, no and unknown.

Should: Null values should have their fields removed

Swagger/OpenAPI, which is in common use, doesn’t support null field values (it does allow omitting that field completely if it is not marked as required). However that doesn’t prevent clients and servers sending and receiving those fields with null values. Also, in some cases null may be a meaningful value - for example, JSON Merge Patch RFC 7382) using null to indicate property deletion.

Should: Empty array values should not be null

Empty array values can unambiguously be represented as the empty list, [].

Should: Enumerations should be represented as Strings

Strings are a reasonable target for values that are by design enumerations.

Should: Date property values should conform to RFC 3399

Use the date and time formats defined by RFC 3339:

  • for "date" use strings matching date-fullyear "-" date-month "-" date-mday, for example: 2015-05-28

  • for "date-time" use strings matching full-date "T" full-time, for example 2015-05-28T14:07:17Z

Note that the OpenAPI format "date-time" corresponds to "date-time" in the RFC) and 2015-05-28 for a date (note that the OpenAPI format "date" corresponds to "full-date" in the RFC). Both are specific profiles, a subset of the international standard ISO 8601.

A zone offset may be used (both, in request and responses) — this is simply defined by the standards. However, we encourage restricting dates to UTC and without offsets. For example 2015-05-28T14:07:17Z rather than 2015-05-28T14:07:17+00:00. From experience we have learned that zone offsets are not easy to understand and often not correctly handled. Note also that zone offsets are different from local times that might be including daylight saving time. Localization of dates should be done by the services that provide user interfaces, if required.

When it comes to storage, all dates should be consistently stored in UTC without a zone offset. Localization should be done locally by the services that provide user interfaces, if required.

Sometimes it can seem data is naturally represented using numerical timestamps, but this can introduce interpretation issues with precision - for example whether to represent a timestamp as 1460062925, 1460062925000 or 1460062925.000. Date strings, though more verbose and requiring more effort to parse, avoid this ambiguity.

May: Time durations and intervals could conform to ISO 8601

Schema based JSON properties that are by design durations and intervals could be strings formatted as recommended by ISO 8601 (Appendix A of RFC 3399 contains a grammar for durations).

May: Standards could be used for Language, Country and Currency

7. API Naming

Must: Use lowercase separate words with hyphens for Path Segments



This applies to concrete path segments and not the names of path parameters. For example {shipment_order_id} would be ok as a path parameter.

Must: Use snake_case (never camelCase) for Query Parameters


customer_number, order_id, billing_address

Should: Prefer Hyphenated-Pascal-Case for HTTP header Fields

This is for consistency in your documentation (most other headers follow this convention). Avoid camelCase (without hyphens). Exceptions are common abbreviations like “ID.”



May: Use Standardized Headers

Use this list and mention its support in your OpenAPI definition.

Must: Pluralize Resource Names

Usually, a collection of resource instances is provided (at least API should be ready here). The special case of a resource singleton is a collection with cardinality 1.

May: Use /api as first Path Segment

In most cases, all resources provided by a service are part of the public API, and therefore should be made available under the root “/” base path. If the service should also support non-public, internal APIs — for specific operational support functions, for example — add “/api” as base path to clearly separate public and non-public API resources.

Must: Avoid Trailing Slashes

The trailing slash must not have specific semantics. Resource paths must deliver the same results whether they have the trailing slash or not.

May: Use Conventional Query Strings

If you provide query support for sorting, pagination, filtering functions or other actions, use the following standardized naming conventions:

  • q — default query parameter (e.g. used by browser tab completion); should have an entity specific alias, like sku

  • limit — to restrict the number of entries. See Pagination section below. Hint: You can use size as an alternate query string.

  • cursor — key-based page start. See Pagination section below.

  • offset — numeric offset page start. See Pagination section below. Hint: In combination with limit, you can use page as an alternative to offset.

  • sort — comma-separated list of fields to sort. To indicate sorting direction, fields my prefixed with + (ascending) or - (descending, default), e.g. /sales-orders?sort=+id

  • fields — to retrieve a subset of fields. See Support Filtering of Resource Fields below.

  • embed — to expand embedded entities (ie.: inside of an article entity, expand silhouette code into the silhouette object). Implementing “expand” correctly is difficult, so do it with care.

    = Resources

Must: Avoid Actions — Think About Resources

REST is all about your resources, so consider the domain entities that take part in web service interaction, and aim to model your API around these using the standard HTTP methods as operation indicators. For instance, if an application has to lock articles explicitly so that only one user may edit them, create an article lock with PUT or POST instead of using a lock action.


PUT /article-locks/{article-id}

The added benefit is that you already have a service for browsing and filtering article locks.

Should: Model complete business processes

An API should contain the complete business processes containing all resources representing the process. This enables clients to understand the business process, foster a consistent design of the business process, allow for synergies from description and implementation perspective, and eliminates implicit invisible dependencies between APIs.

In addition, it prevents services from being designed as thin wrappers around databases, which normally tends to shift business logic to the clients.

Should: Define useful resources

As a rule of thumb resources should be defined to cover 90% of all its client’s use cases. A useful resource should contain as much information as necessary, but as little as possible. A great way to support the last 10% is to allow clients to specify their needs for more/less information by supporting filtering and embedding.

Must: Keep URLs Verb-Free

The API describes resources, so the only place where actions should appear is in the HTTP methods. In URLs, use only nouns. Instead of thinking of actions (verbs), it’s often helpful to think about putting a message in a letter box: e.g., instead of having the verb cancel in the url, think of sending a message to cancel an order to the cancellations letter box on the server side.

Must: Use Domain-Specific Resource Names

API resources represent elements of the application’s domain model. Using domain-specific nomenclature for resource names helps developers to understand the functionality and basic semantics of your resources. It also reduces the need for further documentation outside the API definition. For example, “sales-order-items” is superior to “order-items” in that it clearly indicates which business object it represents. Along these lines, “items” is too general.

Must: Identify resources and Sub-Resources via Path Segments

Some API resources may contain or reference sub-resources. Embedded sub-resources, which are not top-level resources, are parts of a higher-level resource and cannot be used outside of its scope. Sub-resources should be referenced by their name and identifier in the path segments.

Composite identifiers must not contain / as a separator. In order to improve the consumer experience, you should aim for intuitively understandable URLs, where each sub-path is a valid reference to a resource or a set of resources. For example, if /customers/12ev123bv12v/addresses/DE_100100101 is a valid path of your API, then /customers/12ev123bv12v/addresses, /customers/12ev123bv12v and /customers must be valid as well in principle.

Basic URL structure:




Should: Only Use UUIDs If Necessary

Generating IDs can be a scaling problem in high frequency and near real time use cases. UUIDs solve this problem, as they can be generated without collisions in a distributed, non-coordinated way and without additional server roundtrips.

However, they also come with some disadvantages:

  • pure technical key without meaning; not ready for naming or name scope conventions that might be helpful for pragmatic reasons, e.g. we learned to use names for product attributes, instead of UUIDs

  • less usable, because…​

  • cannot be memorized and easily communicated by humans

  • harder to use in debugging and logging analysis

  • less convenient for consumer facing usage

  • quite long: readable representation requires 36 characters and comes with higher memory and bandwidth consumption

  • not ordered along their creation history and no indication of used id volume

  • may be in conflict with additional backward compatibility support of legacy ids

UUIDs should be avoided were not needed for large scale id generation. Instead, for instance, server side support with id generation can be preferred (POST on id resource, followed by idempotent PUT on entity resource). Usage of UUIDs is especially discouraged as primary keys of master and configuration data, like brand-ids or attribute-ids which have low id volume but widespread steering functionality.

In any case, we should always use string rather than number type for identifiers. This gives us more flexibility to evolve the identifier naming scheme. Accordingly, if used as identifiers, UUIDs should not be qualified using a format property.

Hint: Usually, random UUID is used - see UUID version 4 in RFC 4122. Though UUID version 1 also contains leading timestamps it is not reflected by its lexicographic sorting. This deficit is addressed by ULID (Universally Unique Lexicographically Sortable Identifier). You may favour ULID instead of UUID, for instance, for pagination use cases ordered along creation time.

May: Consider Using (Non-) Nested URLs

If a sub-resource is only accessible via its parent resource and may not exists without parent resource, consider using a nested URL structure, for instance:


However, if the resource can be accessed directly via its unique id, then the API should expose it as a top-level resource. For example, customer has a collection for sales orders; however, sales orders have globally unique id and some services may choose to access the orders directly, for instance:


Should: Limit number of Resources

To keep maintenance and service evolution manageable, we should follow "functional segmentation" and "separation of concern" design principles and do not mix different business functionalities in same API definition. In this sense the number of resources exposed via API should be limited - our experience is that a typical range of resources for a well-designed API is between 4 and 8. There may be exceptions with more complex business domains that require more resources, but you should first check if you can split them into separate subdomains with distinct APIs.

Nevertheless one API should hold all necessary resources to model complete business processes helping clients to understand these flows.

Should: Limit number of Sub-Resource Levels

There are main resources (with root url paths) and sub-resources (or “nested” resources with non-root urls paths). Use sub-resources if their life cycle is (loosely) coupled to the main resource, i.e. the main resource works as collection resource of the subresource entities. You should use ⇐ 3 sub-resource (nesting) levels — more levels increase API complexity and url path length. (Remember, some popular web browsers do not support URLs of more than 2000 characters)


Must: Use HTTP Methods Correctly

Be compliant with the standardized HTTP method semantics summarized as follows:


GET requests are used to read a single resource or query set of resources.

  • GET requests for individual resources will usually generate a 404 if the resource does not exist

  • GET requests for collection resources may return either 200 (if the listing is empty) or 404 (if the list is missing)

  • GET requests must NOT have request body payload

Note: GET requests on collection resources should provide a sufficient filter mechanism as well as Pagination.

"GET with Body"

APIs sometimes face the problem, that they have to provide extensive structured request information with GET, that may even conflicts with the size limits of clients, load-balancers, and servers. As we require APIs to be standard conform (body in GET must be ignored on server side), API designers have to check the following two options:

  1. GET with URL encoded query parameters: when it is possible to encode the request information in query parameters, respecting the usual size limits of clients, gateways, and servers, this should be the first choice. The request information can either be provided distributed to multiple query parameters or a single structured URL encoded string.

  2. POST with body content: when a GET with URL encoded query parameters is not possible, a POST with body content must be used. In this case the endpoint must be documented with the hint GET with body to transport the GET semantic of this call.

Note: It is no option to encode the lengthy structured request information in header parameters. From a conceptual point of view, the semantic of an operation should always be expressed by resource name and query parameters, i.e. what goes into the URL. Request headers are reserved for general context information, e.g. FlowIDs. In addition, size limits on query parameters and headers are not reliable and depend on clients, gateways, server, and actual settings. Thus, switching to headers does not solve the original problem.


PUT requests are used to create or update entire resources - single or collection resources. The semantic is best described as »please put the enclosed representation at the resource mentioned by the URL, replacing any existing resource.«.

  • PUT requests are usually applied to single resources, and not to collection resources, as this would imply replacing the entire collection

  • PUT requests are usually robust against non-existence of resources by implicitly creating before updating

  • on successful PUT requests, the server will replace the entire resource addressed by the URL with the representation passed in the payload (subsequent reads will deliver the same payload)

  • successful PUT requests will usually generate 200 or 204 (if the resource was updated - with or without actual content returned), and 201 (if the resource was created)

Note: Resource IDs with respect to PUT requests are maintained by the client and passed as a URL path segment. Putting the same resource twice is required to be idempotent and to result in the same single resource instance. If PUT is applied for creating a resource, only URIs should be allowed as resource IDs. If URIs are not available POST should be preferred.

To prevent unnoticed concurrent updates when using PUT, the combination of ETag and If-(None-)Match headers should be considered to signal the server stricter demands to expose conflicts and prevent lost updates.


POST requests are idiomatically used to create single resources on a collection resource endpoint, but other semantics on single resources endpoint are equally possible. The semantic for collection endpoints is best described as »please add the enclosed representation to the collection resource identified by the URL«. The semantic for single resource endpoints is best described as »please execute the given well specified request on the collection resource identified by the URL«.

  • POST request should only be applied to collection resources, and normally not on single resource, as this has an undefined semantic

  • on successful POST requests, the server will create one or multiple new resources and provide their URI/URLs in the response

  • successful POST requests will usually generate 200 (if resources have been updated), 201 (if resources have been created), and 202 (if the request was accepted but has not been finished yet)

More generally: POST should be used for scenarios that cannot be covered by the other methods sufficiently. For instance, GET with complex (e.g. SQL like structured) query that needs to be passed as request body payload because of the URL-length constraint. In such cases, make sure to document the fact that POST is used as a workaround.

Note: Resource IDs with respect to POST requests are created and maintained by server and returned with response payload. Posting the same resource twice is by itself not required to be idempotent and may result in multiple resource instances. Anyhow, if external URIs are present that can be used to identify duplicate requests, it is best practice to implement POST in an idempotent way.


PATCH request are only used for partial update of single resources, i.e. where only a specific subset of resource fields should be replaced. The semantic is best described as »please change the resource identified by the URL according to my change request«. The semantic of the change request is not defined in the HTTP standard and must be described in the API specification by using suitable media types.

  • PATCH requests are usually applied to single resources, and not on collection resources, as this would imply patching on the entire collection

  • PATCH requests are usually not robust against non-existence of resource instances

  • on successful PATCH requests, the server will update parts of the resource addressed by the URL as defined by the change request in the payload

  • successful PATCH requests will usually generate 200 or 204 (if resources have been updated

  • with or without updated content returned)

Note: since implementing PATCH correctly is a bit tricky, we strongly suggest to choose one and only one of the following patterns per endpoint, unless forced by a backwards compatible change. In preference order:

  1. use PUT with complete objects to update a resource as long as feasible (i.e. do not use PATCH at all).

  2. use PATCH with partial objects to only update parts of a resource, when ever possible. (This is basically JSON Merge Patch, a specialized media type application/merge-patch+json that is a partial resource representation.)

  3. use PATCH with JSON Patch, a specialized media type application/json-patch+json that includes instructions on how to change the resource.

  4. use POST (with a proper description of what is happening) instead of PATCH if the request does not modify the resource in a way defined by the semantics of the media type.

In practice JSON Merge Patch quickly turns out to be too limited, especially when trying to update single objects in large collections (as part of the resource). In this cases JSON Patch can shown its full power while still showing readable patch requests (see also JSON patch vs. merge).

To prevent unnoticed concurrent updates when using PATCH, the combination of ETag`and `If-Match headers should be considered to signal the server stricter demands to expose conflicts and prevent lost updates.


DELETE request are used to delete resources. The semantic is best described as »please delete the resource identified by the URL«.

  • DELETE requests are usually applied to single resources, not on collection resources, as this would imply deleting the entire collection

  • successful DELETE request will usually generate 200 (if the deleted resource is returned) or 204 (if no content is returned)

  • failed DELETE request will usually generate 404 (if the resource cannot be found) or 410 (if the resource was already deleted before)

HEAD requests are used retrieve to header information of single resources and resource collections.

  • HEAD has exactly the same semantics as GET, but returns headers only, no body.


OPTIONS are used to inspect the available operations (HTTP methods) of a given endpoint.

  • OPTIONS requests usually either return a comma separated list of methods (provided by an Allow:-Header) or as a structured list of link templates

Note: OPTIONS is rarely implemented, though it could be used to self-describe the full functionality of a resource.

Must: Fulfill Safeness and Idempotency Properties

An operation can be…​

  • idempotent, i.e. operation will produce the same results if executed once or multiple times (note: this does not necessarily mean returning the same status code)

  • safe, i.e. must not have side effects such as state changes

Method implementations must fulfill the following basic properties:

HTTP method safe idempotent






















Please see also Best Practices [internal link] for further hints on how to support the different HTTP methods on resources.

Must: Use Specific HTTP Status Codes

This guideline groups the following rules for HTTP status codes usage:

  • You must not invent new HTTP status codes; only use standardized HTTP status codes and consistent with its intended semantics.

  • You should use the most specific HTTP status code for your concrete resource request processing status or error situation.

  • You should provide good documentation in the API definition when using HTTP status codes that are less commonly used and not listed below.

There are \~60 different HTTP status codes with specific semantics defined in the HTTP standards (mainly RFC7231 and RFC-6585) - and there are upcoming new ones, e.g. draft legally-restricted-status (see overview on all error codes on Wikipedia or via https://httpstatuses.com/). And there are unofficial ones, e.g. used by specific web servers like Nginx.

Our list of most commonly used and best understood HTTP status codes:

Success Codes

Code Meaning Methods


OK - this is the standard success response



Created - Returned on successful entity creation. You are free to return either an empty response or the created resource in conjunction with the Location header. (More details found in the Common Headers.) Always set the Location header.



Accepted - The request was successful and will be processed asynchronously.



No content - There is no response body



Multi-Status - The response body contains multiple status informations for different parts of a batch/bulk request. See Must: Provide Error Documentation.


Redirection Codes

Code Meaning Methods


Moved Permanently - This and all future requests should be directed to the given URI.



See Other - The response to the request can be found under another URI using a GET method.



Not Modified - resource has not been modified since the date or version passed via request headers If-Modified-Since or If-None-Match.


Client Side Error Codes

Code Meaning Methods


Bad request - generic / unknown error



Unauthorized - the users must log in (this often means “Unauthenticated”)



Forbidden - the user is not authorized to use this resource



Not found - the resource is not found



Method Not Allowed - the method is not supported, see OPTIONS



Not Acceptable - resource can only generate content not acceptable according to the Accept headers sent in the request



Request timeout - the server times out waiting for the resource



Conflict - request cannot be completed due to conflict, e.g. when two clients try to create the same resource or if there are concurrent, conflicting updates



Gone - resource does not exist any longer, e.g. when accessing a resource that has intentionally been deleted



Precondition Failed - returned for conditional requests, e.g. If-Match if the condition failed. Used for optimistic locking.



Unsupported Media Type - e.g. clients sends request body without content type



Locked - Pessimistic locking, e.g. processing states



Precondition Required - server requires the request to be conditional (e.g. to make sure that the “lost update problem” is avoided).



Too many requests - the client does not consider rate limiting and sent too many requests. See Must: Use 207 for Batch or Bulk Requests.


Server Side Error Codes:

Code Meaning Methods


Internal Server Error - a generic error indication for an unexpected server execution problem (here, client retry may be senseful)



Not Implemented - server cannot fulfill the request (usually implies future availability, e.g. new feature).



Service Unavailable - server is (temporarily) not available (e.g. due to overload) — client retry may be senseful.


Must: Provide Error Documentation

APIs should define the functional, business view and abstract from implementation aspects. Errors become a key element providing context and visibility into how to use an API. The error object should be extended by an application-specific error identifier if and only if the HTTP status code is not specific enough to convey the domain-specific error semantic. For this reason, we use a standardized error return object definition — see Must: Follow Hypertext Control Conventions.

The OpenAPI specification shall include definitions for error descriptions that will be returned; they are part of the interface definition and provide important information for service clients to handle exceptional situations and support troubleshooting. You should also think about a troubleshooting board — it is part of the associated online API documentation, provides information and handling guidance on application-specific errors and is referenced via links of the API definition. This can reduce service support tasks and contribute to service client and provider performance.

Service providers should differentiate between technical and functional errors. In most cases it’s not useful to document technical errors that are not in control of the service provider unless the status code convey application-specific semantics. The list of status code that can be omitted from API specifications includes but is not limited to:

  • 401 Unauthorized

  • 403 Forbidden

  • 404 Not Found unless it has some additional semantics

  • 405 Method Not Allowed

  • 406 Not Acceptable

  • 408 Request Timeout

  • 413 Payload Too Large

  • 414 URI Too Long

  • 415 Unsupported Media Type

  • 500 Internal Server Error

  • 502 Bad Gateway

  • 503 Service Unavailable

  • 504 Gateway Timeout

Even though they might not be documented - they may very much occur in production, so clients should be prepared for unexpected response codes, and in case of doubt handle them like they would handle the corresponding x00 code. Adding new response codes (specially error responses) should be considered a compatible API evolution.

Functional errors on the other hand, that convey domain-specific semantics, must be documented and are strongly encouraged to be expressed with Must: Follow Hypertext Control Conventions.

Must: Use 207 for Batch or Bulk Requests

Some APIs are required to provide either batch or bulk requests using POST for performance reasons, i.e. for communication and processing efficiency. In this case services may be in need to signal multiple response codes for each part of an batch or bulk request. As HTTP does not provide proper guidance for handling batch/bulk requests and responses, we herewith define the following approach:

  • A batch or bulk request always has to respond with HTTP status code 207, unless it encounters a generic or unexpected failure before looking at individual parts.

  • A batch or bulk response with status code 207 always returns a multi-status object containing sufficient status and/or monitoring information for each part of the batch or bulk request.

  • A batch or bulk request may result in a status code 400/500, only if the service encounters a failure before looking at individual parts or, if an unanticipated failure occurs.

The before rules apply even in the case that processing of all individual part fail or each part is executed asynchronously! They are intended to allow clients to act on batch and bulk responses by inspecting the individual results in a consistent way.

Note: while a batch defines a collection of requests triggering independent processes, a bulk defines a collection of independent resources created or updated together in one request. With respect to response processing this distinction normally does not matter.

Must: Use 429 with Headers for Rate Limits

APIs that wish to manage the request rate of clients must use the '429 Too Many Requests' response code if the client exceeded the request rate and therefore the request can’t be fulfilled. Such responses must also contain header information providing further details to the client. There are two approaches a service can take for header information:

  • Return a 'Retry-After' header indicating how long the client ought to wait before making a follow-up request. The Retry-After header can contain a HTTP date value to retry after or the number of seconds to delay. Either is acceptable but APIs should prefer to use a delay in seconds.

  • Return a trio of 'X-RateLimit' headers. These headers (described below) allow a server to express a service level in the form of a number of allowing requests within a given window of time and when the window is reset.

The 'X-RateLimit' headers are:

  • X-RateLimit-Limit: The maximum number of requests that the client is allowed to make in this window.

  • X-RateLimit-Remaining: The number of requests allowed in the current window.

  • X-RateLimit-Reset: The relative time in seconds when the rate limit window will be reset.

The reason to allow both approaches is that APIs can have different needs. Retry-After is often sufficient for general load handling and request throttling scenarios and notably, does not strictly require the concept of a calling entity such as a tenant or named account. In turn this allows resource owners to minimise the amount of state they have to carry with respect to client requests. The 'X-RateLimit' headers are suitable for scenarios where clients are associated with pre-existing account or tenancy structures. 'X-RateLimit' headers are generally returned on every request and not just on a 429, which implies the service implementing the API is carrying sufficient state to track the number of requests made within a given window for each named entity.

Should: Explicitly define the Collection Format of Query Parameters

There are different ways of supplying a set of values as a query parameter. One particular type should be selected and stated explicitly in the API definition. The OpenAPI property collectionFormat is used to specify the format of the query parameter.

Only the csv or multi formats should be used for multi-value query parameters as described below.

Collection Format Description Example


Comma separated values



Multiple parameter instances


When choosing the collection format, take into account the tool support, the escaping of special characters and the maximal URL length.

9. Performance

Should: Reduce Bandwidth Needs and Improve Responsiveness

APIs should support techniques for reducing bandwidth based on client needs. This holds for APIs that (might) have high payloads and/or are used in high-traffic scenarios like the public Internet and telecommunication networks. Typical examples are APIs used by mobile web app clients with (often) less bandwidth connectivity. (Zalando is a 'Mobile First' company, so be mindful of this point.)

Common techniques include:

Each of these items is described in greater detail below.

Should: Use gzip Compression

Compress the payload of your API’s responses with gzip, unless there’s a good reason not to — for example, you are serving so many requests that the time to compress becomes a bottleneck. This helps to transport data faster over the network (fewer bytes) and makes frontends respond faster.

Though gzip compression might be the default choice for server payload, the server should also support payload without compression and its client control via Accept-Encoding request header — see also RFC 7231 Section 5.3.4. The server should indicate used gzip compression via the Content-Encoding header.

Should: Support Filtering of Resource Fields

Depending on your use case and payload size, you can significantly reduce network bandwidth need by supporting filtering of returned entity fields. Here, the client can determine the subset of fields he wants to receive via the fields query parameter — example see Google AppEngine API’s partial response:


GET http://api.example.org/resources/123 HTTP/1.1

HTTP/1.1 200 OK
Content-Type: application/json

  "id": "cddd5e44-dae0-11e5-8c01-63ed66ab2da5",
  "name": "John Doe",
  "address": "1600 Pennsylvania Avenue Northwest, Washington, DC, United States",
  "birthday": "1984-09-13",
  "partner": {
    "id": "1fb43648-dae1-11e5-aa01-1fbc3abb1cd0",
    "name": "Jane Doe",
    "address": "1600 Pennsylvania Avenue Northwest, Washington, DC, United States",
    "birthday": "1988-04-07"


GET http://api.example.org/resources/123?fields=(name,partner(name)) HTTP/1.1

HTTP/1.1 200 OK
Content-Type: application/json

  "name": "John Doe",
  "partner": {
    "name": "Jane Doe"

As illustrated by this example, field filtering should be done via request parameter "fields" with value range defined by the following BNF grammar.

<fields> ::= <negation> <fields_expression> | <fields_expression>

<negation> ::= "!"

<fields_expression> ::= "(" <field_set> ")"

<field_set> ::= <qualified_field> | <qualified_field> "," <field_set>

<qualified_field> ::= <field> | <field> <fields_expression>

<field> ::= <DASH_LETTER_DIGIT> | <DASH_LETTER_DIGIT> <field>


<DASH> ::= "-" | "_"

<LETTER> ::= "A" | "B" | "C" | "D" | "E" | "F" | "G" | "H" | "I" | "J" | "K" | "L" | "M" | "N" | "O" | "P" | "Q" | "R" | "S" | "T" | "U" | "V" | "W" | "X" | "Y" | "Z" | "a" | "b" | "c" | "d" | "e" | "f" | "g" | "h" | "i" | "j" | "k" | "l" | "m" | "n" | "o" | "p" | "q" | "r" | "s" | "t" | "u" | "v" | "w" | "x" | "y" | "z"

<DIGIT> ::= "0" | "1" | "2" | "3" | "4" | "5" | "6" | "7" | "8" | "9"

A fields_expression as defined by the grammar describes the properties of an object, i.e. (name) returns only the name property of the root object. (name,partner(name)) returns the name and partner properties where partner itself is also an object and only its name property is returned.

Hint: OpenAPI doesn’t allow you to formally specify whether depending on a given parameter will return different parts of the specified result schema. Explain this in English in the parameter description.

Should: Allow Optional Embedding of Sub-Resources

Embedding related resources (also know as Resource expansion) is a great way to reduce the number of requests. In cases where clients know upfront that they need some related resources they can instruct the server to prefetch that data eagerly. Whether this is optimized on the server, e.g. a database join, or done in a generic way, e.g. an HTTP proxy that transparently embeds resources, is up to the implementation.

See Must: Avoid Trailing Slashes for naming, e.g. "embed" for steering of embedded resource expansion. Please use the BNF grammar, as already defined above for filtering, when it comes to an embedding query syntax.

Embedding a sub-resource can possibly look like this where an order resource has its order items as sub-resource (/order/{orderId}/items):

GET /order/123?embed=(items) HTTP/1.1

  "id": "123",
  "_embedded": {
    "items": [
        "position": 1,
        "sku": "1234-ABCD-7890",
        "price": {
          "amount": 71.99,
          "currency": "EUR"

10. Pagination

Must: Support Pagination

Access to lists of data items must support pagination for best client side batch processing and iteration experience. This holds true for all lists that are (potentially) larger than just a few hundred entries.

There are two page iteration techniques:

The technical conception of pagination should also consider user experience related issues. As mentioned in this article, jumping to a specific page is far less used than navigation via next/previous page links. This favours cursor-based over offset-based pagination.

Should: Prefer Cursor-Based Pagination, Avoid Offset-Based Pagination

Cursor-based pagination is usually better and more efficient when compared to offset-based pagination. Especially when it comes to high-data volumes and / or storage in NoSQL databases.

Before choosing cursor-based pagination, consider the following trade-offs:

  • Usability/framework support:

    • Offset / limit based pagination is more known than cursor-based pagination, so it has more framework support and is easier to use for API clients

  • Use case: Jump to a certain page

    • If jumping to a particular page in a range (e.g., 51 of 100) is really a required use case, cursor-based navigation is not feasible

  • Variability of data may lead to anomalies in result pages

    • Offset-based pagination may create duplicates or lead to missing entries if rows are inserted or deleted between two subsequent paging requests.

    • When using cursor-based pagination, paging cannot continue when the cursor entry has been deleted while fetching two pages

  • Performance considerations - efficient server-side processing using offset-based pagination is hardly feasible for:

    • Higher data list volumes, especially if they do not reside in the database’s main memory

    • Sharded or NoSQL databases

  • Cursor-based navigation may not work if you need the total count of results and / or backward iteration support

Further reading:

May: Use Pagination Links Where Applicable

Those collections should then have an items attribute holding the items of the current page. The collection may contain additional metadata about the collection or the current page (e.g. index, page_size) when necessary.

You should avoid providing a total count in your API unless there’s a clear need to do so. Very often, there are systems and performance implications to supporting full counts, especially as datasets grow and requests become complex queries or filters that drive full scans (e.g., your database might need to look at all candidate items to count them). While this is an implementation detail relative to the API, it’s important to consider your ability to support serving counts over the life of a service.

If the collection consists of links to other resources, the collection name should use IANA registered link relations as names whenever appropriate, but use plural form.

E.g. a service for articles could represent the collection of hyperlinks to an article’s authors like that:

  "self": "https://.../articles/xyz/authors/",
  "index": 0,
  "page_size": 5,
  "items": [
      "href": "https://...",
      "id": "123e4567-e89b-12d3-a456-426655440000",
      "name": "Kent Beck"
      "href": "https://...",
      "id": "987e2343-e89b-12d3-a456-426655440000",
      "name": "Mike Beedle"
  "first": "https://...",
  "next": "https://...",
  "prev": "https://...",
  "last": "https://..."

11. Hypermedia

Must: Use REST Maturity Level 2

We strive for a good implementation of REST Maturity Level 2 as it enables us to build resource-oriented APIs that make full use of HTTP verbs and status codes. You can see this expressed by many rules throughout these guidelines, e.g.:

Although this is not HATEOAS, it should not prevent you from designing proper link relationships in your APIs as stated in rules below.

May: Use REST Maturity Level 3 - HATEOAS

We do not generally recommend to implement REST Maturity Level 3. HATEOAS comes with additional API complexity without real value in our SOA context where client and server interact via REST APIs and provide complex business functions as part of our e-commerce SaaS platform.

Our major concerns regarding the promised advantages of HATEOAS (see also RESTistential Crisis over Hypermedia APIs, Why I Hate HATEOAS and others): * We follow the API First principle with APIs explicitly defined outside the code with standard specification language. HATEOAS does not really add value for SOA client engineers in terms of API self-descriptiveness: a client engineer finds necessary links and usage description (depending on resource state) in the API reference definition anyway. * Generic HATEOAS clients which need no prior knowledge about APIs and explore API capabilities based on hypermedia information provided, is a theoretical concept that we haven’t seen working in practise and does not fit to our SOA set-up. The OpenAPI description format (and tooling based on OpenAPI) doesn’t provide sufficient support for HATEOAS either. * In practice relevant HATEOAS approximations (e.g. following specifications like HAL or JSON API) support API navigation by abstracting from URL endpoint and HTTP method aspects via link types. So, Hypermedia does not prevent clients from required manual changes when domain model changes over time. * Hypermedia make sense for humans, less for SOA machine clients. We would expect use cases where it may provide value more likely in the frontend and human facing service domain. * Hypermedia does not prevent API clients to implement shortcuts and directly target resources without 'discovering' them.

However, we do not forbid HATEOAS; you could use it, if you checked its limitations and still see clear value for your usage scenario that justifies its additional complexity. If you use HATEOAS please share experience and present your findings in the API Guild [internal link].

Must: Use Common Hypertext Controls

When embedding links to other resources into representations you must use the common hypertext control object. It contains at least one attribute:

  • href: The URI of the resource the hypertext control is linking to. All our API are using HTTP(s) as URI scheme.

In API that contain any hypertext controls, the attribute name href is reserved for usage within hypertext controls.

The schema for hypertext controls can be derived from this model:

  description: A base type of objects representing links to resources.
  type: object
      description: Any URI that is using http or https protocol
      type: string
      format: uri
  required: [ "href" ]

The name of an attribute holding such a HttpLink object specifies the relation between the object that contains the link and the linked resource. Implementations should use names from the IANA Link Relation Registry whenever appropriate. As IANA link relation names use hyphen-case notation, while this guide enforces snake_case notation for attribute names, hyphens in IANA names have to be replaced with underscores (e.g. the IANA link relation type version-history would become the attribute version_history)

Specific link objects may extend the basic link type with additional attributes, to give additional information related to the linked resource or the relationship between the source resource and the linked one.

E.g. a service providing "Person" resources could model a person who is married with some other person with a hypertext control that contains attributes which describe the other person (id, name) but also the relationship "spouse" between between the two persons (since):

  "id": "446f9876-e89b-12d3-a456-426655440000",
  "name": "Peter Mustermann",
  "spouse": {
    "href": "https://...",
    "since": "1996-12-19",
    "id": "123e4567-e89b-12d3-a456-426655440000",
    "name": "Linda Mustermann"

Hypertext controls are allowed anywhere within a JSON model. While this specification would allow HAL, we actually don’t recommend/enforce the usage of HAL anymore as the structural separation of meta-data and data creates more harm than value to the understandability and usability of an API.

Should: Use Simple Hypertext Controls for Pagination and Self-References

Hypertext controls for pagination inside collections and self-references should use a simple URI value in combination with their corresponding link relations (next, prev, first, last, self) instead of the extensible common hypertext control

See Pagination for information how to best represent paginateable collections.

Must: Not Use Link Headers with JSON entities

We don’t allow the use of the Link Header defined by RFC 5988 in conjunction with JSON media types. We prefer links directly embedded in JSON payloads to the uncommon link header syntax.

12. Data Formats

Must: Use JSON to Encode Structured Data

Use JSON-encoded body payload for transferring structured data. The JSON payload must follow RFC-7159 by having (if possible) a serialized object as the top-level structure, since it would allow for future extension. This also applies for collection resources where one naturally would assume an array. See Should: Prefer Cursor-Based Pagination, Avoid Offset-Based Pagination for an example.

May: Use non JSON Media Types for Binary Data or Alternative Content Representations

Other media types may be used in following cases:

  • Transferring binary data or data whose structure is not relevant. This is the case if payload structure is not interpreted and consumed by clients as is. Example of such use case is downloading images in formats JPG, PNG, GIF.

  • In addition to JSON version alternative data representations (e.g. in formats PDF, DOC, XML) may be made available through content negotiation.

Must: Use Standard Date and Time Formats

JSON Payload

Read more about date and time format in Should: Enumerations should be represented as Strings.

HTTP headers

Http headers including the proprietary headers use the HTTP date format defined in RFC 7231.

May: Use Standards for Country, Language and Currency Codes

Use the following standard formats for country, language and currency codes:

Must: Define Format for Type Number and Integer

Whenever an API defines a property of type number or integer, the precision must be defined by the format as follows to prevent clients from guessing the precision incorrectly, and thereby changing the value unintentionally:

type format specified value range



integer between -231 and 231-1



integer between -263 and 263-1



arbitrarily large signed integer number



IEEE 754-2008/ISO 60559:2011 binary64 decimal number



IEEE 754-2008/ISO 60559:2011 binary128 decimal number



arbitrarily precise signed decimal number

The precision must be translated by clients and servers into the most specific language types. E.g. for the following definitions the most specific language types in Java will translate to BigDecimal for Money.amount and int or Integer for the OrderList.page_size:

  type: object
      type: number
      description: Amount expressed as a decimal number of major currency units
      format: decimal
      example: 99.95

  type: object
      type: integer
      description: Number of orders in list
      format: int32
      example: 42

Should: Prefer standard Media type name application/json

Previously, this guideline allowed the use of custom media types like application/x.zalando.article+json. This usage is not recommended anymore and should be avoided, except where it is necessary for cases of media type versioning. Instead, the standard media type name application/json (or Must: Follow Hypertext Control Conventions).

Custom media types with subtypes beginning with x bring no advantage compared to the standard media type for JSON, and make automated processing more difficult. They are also discouraged by RFC 6838.

13. Common Data Types

Definitions of data objects that are good candidates for wider usage:

Should: Use a Common Money Object

Use the following common money structure:

  type: object
      type: number
      description: Amount expressed as a decimal number of major currency units
      format: decimal
      example: 99.95
      type: string
      description: 3 letter currency code as defined by ISO-4217
      format: iso-4217
      example: EUR
    - amount
    - currency

The decimal values for "amount" describe unit and subunit of the currency in a single value, where the digits before the decimal point are for the major unit and the digits after the decimal point are for the minor unit. Note that some business cases (e.g. transactions in Bitcoin) call for a higher precision, so applications must be prepared to accept values with unlimited precision, unless explicitly stated otherwise in the API specification. Examples for correct representations (in EUR):

  • 42.20 or 42.2 = 42 Euros, 20 Cent

  • 0.23 = 23 Cent

  • 42.0 or 42 = 42 Euros

  • 1024.42 = 1024 Euros, 42 Cent

  • 1024.4225 = 1024 Euros, 42.25 Cent

Make sure that you don’t convert the “amount” field to float / double types when implementing this interface in a specific language or when doing calculations. Otherwise, you might lose precision. Instead, use exact formats like Java’s BigDecimal. See Stack Overflow for more info.

Some JSON parsers (NodeJS’s, for example) convert numbers to floats by default. After discussing the pros and cons (internal link), we’ve decided on "decimal" as our amount format. It is not a standard OpenAPI format, but should help us to avoid parsing numbers as float / doubles.

Must: Use common field names and semantics

There exist a variety of field types that are required in multiple places. To achieve consistency across all API implementations, you must use common field names and semantics whenever applicable.

Generic Fields

There are some data fields that come up again and again in API data:

  • id: the identity of the object. If used, IDs must opaque strings and not numbers. IDs are unique within some documented context, are stable and don’t change for a given object once assigned, and are never recycled cross entities.

  • xyz_id: an attribute within one object holding the identifier of another object must use a name that corresponds to the type of the referenced object or the relationship to the referenced object followed by _id (e.g. customer_id not customer_number; parent_node_id for the reference to a parent node from a child node, even if both have the type Node)

  • created: when the object was created. If used, this must be a date-time construct.

  • modified: when the object was updated. If used, this must be a date-time construct.

  • type: the kind of thing this object is. If used, the type of this field should be a string. Types allow runtime information on the entity provided that otherwise requires examining the Open API file.

Example JSON schema:

  type: object
      description: the identifier of this node
      type: string
      description: when got this node created
      type: string
      format: 'date-time'
      description: when got this node last updated
      type: string
      format: 'date-time'
      type: string
      enum: [ 'LEAF', 'NODE' ]
      description: the identifier of the parent node of this node
      type: string
    id: '123435'
    created: '2017-04-12T23:20:50.52Z'
    modified: '2017-04-12T23:20:50.52Z'
    type: 'LEAF'
    parent_node_id: '534321'

These properties are not always strictly necessary, but making them idiomatic allows API client developers to build up a common understanding of Zalando’s resources. There is very little utility for API consumers in having different names or value types for these fields across APIs.

Address Fields

Address structures play a role in different functional and use-case contexts, including country variances. All attributes that relate to address information should follow the naming and semantics defined below.

  description: a (natural or legal) person that gets addressed
  type: object
    - first_name
    - last_name
    - street
    - city
    - zip
    - country_code
      description: |
        a salutation and/or title used for personal contacts to some
        addressee; not to be confused with the gender information!
      type: string
      example: Mr
      description: |
        given name(s) or first name(s) of a person; may also include the
        middle names.
      type: string
      example: Hans Dieter
      description: |
        family name(s) or surname(s) of a person
      type: string
      example: Mustermann
      description: |
        company name of the business organization. Used when a business is
        the actual addressee; for personal shipments to office addresses, use
        `care_of` instead.
      type: string
      example: Consulting Services GmbH
    - first_name
    - last_name

    an address of a location/destination
  type: object
      description: |
        (aka c/o) the person that resides at the address, if different from
        addressee. E.g. used when sending a personal parcel to the
        office /someone else's home where the addressee resides temporarily
      type: string
      example: Consulting Services GmbH
      description: |
        the full street address including house number and street name
      type: string
      example: Schönhauser Allee 103
      description: |
        further details like building name, suite, apartment number, etc.
      type: string
      example: 2. Hinterhof rechts
      description: |
        name of the city / locality
      type: string
      example: Berlin
      description: |
        zip code or postal code
      type: string
      example: 14265
      description: |
        the country code according to
      type: string
      example: DE
    - street
    - city
    - zip
    - country_code

Grouping and cardinality of fields in specific data types may vary based on the specific use case (e.g. combining addressee and address fields into a single type when modeling an address label vs distinct addressee and address types when modeling users and their addresses).

Must: Follow Hypertext Control Conventions

APIs that provide hypertext controls (links) to interconnect API resources must follow the conventions for naming and modeling of hypertext controls as defined in section Hypermedia.

Must: Use Problem JSON

RFC 7807 defines the media type application/problem+json. Operations should return that (together with a suitable status code) when any problem occurred during processing and you can give more details than the status code itself can supply, whether it be caused by the client or the server (i.e. both for 4xx or 5xx errors).

A previous version of this guideline (before the publication of that RFC and the registration of the media type) told to return application/x.problem+json in these cases (with the same contents). Servers for APIs defined before this change should pay attention to the Accept header sent by the client and set the Content-Type header of the problem response correspondingly. Clients of such APIs should accept both media types.

APIs may define custom problems types with extension properties, according to their specific needs.

The Open API schema definition can be found on github. You can reference it by using:

    description: Service Unavailable
      $ref: 'https://zalando.github.io/problem/schema.yaml#/Problem'

Must: Do not expose Stack Traces

Stack traces contain implementation details that are not part of an API, and on which clients should never rely. Moreover, stack traces can leak sensitive information that partners and third parties are not allowed to receive and may disclose insights about vulnerabilities to attackers.

14. Common Headers

This section describes a handful of headers, which we found raised the most questions in our daily usage, or which are useful in particular circumstances but not widely known.

Must: Use Content Headers Correctly

Content or entity headers are headers with a Content- prefix. They describe the content of the body of the message and they can be used in both, HTTP requests and responses. Commonly used content headers include but are not limited to:

May: Use Content-Location Header

The Content-Location header is optional and can be used in successful write operations (PUT, POST or PATCH) or read operations (GET, HEAD) to guide caching and signal a receiver the actual location of the resource transmitted in the response body. This allows clients to identify the resource and to update their local copy when receiving a response with this header.

The Content-Location header can be used to support the following use cases:

  • For reading operations GET and HEAD, a different location than the requested URI can be used to indicate that the returned resource is subject to content negotiations, and that the value provides a more specific identifier of the resource.

  • For writing operations PUT and PATCH, an identical location to the requested URI, can be used to explicitly indicate that the returned resource is the current representation of the newly created or updated resource.

  • For writing operations POST and DELETE, a content location can be used to indicate that the body contains a status report resource in response to the requested action, which is available at provided location.

Note: When using the Content-Location header, the Content-Type header has to be set as well. For example:

GET /products/123/images HTTP/1.1

HTTP/1.1 200 OK
Content-Type: image/png
Content-Location: /products/123/images?format=raw

Should: Use Location Header instead of Content-Location Header

As the correct usage of Content-Location with respect to semantics and caching is difficult, we discourage the use of Content-Location. In most cases it is sufficient to direct clients to the resource location by using the Location header instead without hitting the Content-Location specific ambiguities and complexities.

More details in RFC 7231 7.1.2 Location, Content-Location

May: Use the Prefer header to indicate processing preferences

The Prefer header defined in RFC7240 allows clients to request processing behaviors from servers. RFC7240 pre-defines a number of preferences and is extensible, to allow others to be defined. Support for the Prefer header is entirely optional and at the discretion of API designers, but as an existing Internet Standard, is recommended over defining proprietary "X-" headers for processing directives.

The Prefer header can defined like this in an API definition:

  name: Prefer
  description: |
    The RFC7240 Prefer header indicates that particular server
    behaviors are preferred by the client but are not required
    for successful completion of the request.
    # (indicate the preferences supported by the API)

  in: header
  type: string
  required: false

Supporting APIs may return the Preference-Applied header also defined in RFC7240 to indicate whether the preference was applied.

May: Consider using ETag together with If-(None-)Match header

When creating or updating resources it may be necessary to expose conflicts and to prevent the lost update problem. This can be best accomplished by using the ETag header together with the If-Match and If-None-Match. The contents of an ETag: <entity-tag> header is either (a) a hash of the response body, (b) a hash of the last modified field of the entity, or (c) a version number or identifier of the entity version.

To expose conflicts between concurrent update operations via PUT, POST, or PATCH, the If-Match: <entity-tag> header can be used to force the server to check whether the version of the updated entity is conforming to the requested <entity-tag>. If no matching entity is found, the operation is supposed a to respond with status code 412 - precondition failed.

Beside other use cases, the If-None-Match: header with parameter * can be used in a similar way to expose conflicts in resource creation. If any matching entity is found, the operation is supposed a to respond with status code 412 - precondition failed.

The ETag, If-Match, and If-None-Match headers can be defined as follows in the API definition:

  name: Etag
  description: |
    The RFC7232 ETag header field in a response provides the current entity-
    tag for the selected resource. An entity-tag is an opaque identifier for
    different versionsof a resource over time, regardless whether multiple
    versions are valid at the same time. An entity-tag consists of an opaque
    quoted string, possibly prefixed by a weakness indicator.

  in: header
  type: string
  required: false
  example: W/"xy", "5", "7da7a728-f910-11e6-942a-68f728c1ba70"

  name: If-Match
  description: |
    The RFC7232 If-Match header field in a request requires the server to
    only operate on the resource that matches at least one of the provided
    entity-tags. This allows clients express a precondition that prevent
    the method from being applied, if there have been any changes to the

  in: header
  type: string
  required: false
  example:  "5", "7da7a728-f910-11e6-942a-68f728c1ba70"

  name: If-None-Match
  description: |
    The RFC7232 If-None-Match header field in a request requires the server
    to only operate on the resource if it does not match any of the provided
    entity-tags. If the provided entity-tag is `*`, it is required that the
    resource does not exist at all.

  in: header
  type: string
  required: false
  example: "7da7a728-f910-11e6-942a-68f728c1ba70", *

15. Proprietary Headers

This section shares definitions of proprietary headers that should be named consistently because they address overarching service-related concerns. Whether services support these concerns or not is optional; therefore, the OpenAPI API specification is the right place to make this explicitly visible. Use the parameter definitions of the resource HTTP methods.

Must: Use Only the Specified Proprietary Zalando Headers

As a general rule, proprietary HTTP headers should be avoided. Still they can be useful in cases where context needs to be passed through multiple services in an end-to-end fashion. As such, a valid use-case for a proprietary header is providing context information, which is not a part of the actual API, but is needed by subsequent communication.

From a conceptual point of view, the semantics and intent of an operation should always be expressed by URLs path and query parameters, the method, and the content. Headers are more often used to implement functions close to the protocol considerations, such as flow control, content negotiation, and authentication. Thus, headers are reserved for general context information (RFC-7231).

X- headers were initially reserved for unstandardized parameters, but the usage of X- headers is deprecated (RFC-6648). This complicates the contract definition between consumer and producer of an API following these guidelines, since there is no aligned way of using those headers. Because of this, the guidelines restrict which X- headers can be used and how they are used.

The Internet Engineering Task Force’s states in RFC-6648 that company specific header' names should incorporate the organization’s name. We aim for backward compatibility, and therefore keep the X- prefix.

The following proprietary headers have been specified by this guideline for usage so far. Remember that HTTP header field names are not case-sensitive.

Header field name Type Description Header field value example



The flow id of the request, which is written into the logs and passed to called services. Helpful for operational troubleshooting and log analysis. It supports traceability of requests and identifying request flows through system of many services. It should be a string consisting of just printable ASCII characters (i.e. without whitespace). Verify in a received request that it fits to a specific format, has a sensible maximum length and possibly throw out or escape characters/bytes which could crash your log parsing (line breaks, tabs, spaces, NULL). If a legacy subsystem can only work with flow IDs of a specific format, it needs to define this in its API, or make its own ones.




The tenant id for future platform multitenancy support. Should not be used unless new platform multitenancy is truly supported. But should be used by New Platform Prototyping services. Must be validated for external retailer, supplier, etc. tenant users via OAuth2; details in clarification. Currently only used by New Platform Prototyping services.




Sales channels are owned by retailers and represent a specific consumer segment being addressed with a specific product assortment that is offered via CFA retailer catalogs to consumers (see platform glossary [internal link])




Consumer facing applications (CFAs) provide business experience to their customers via different frontend application types, for instance, mobile app or browser. Info should be passed-through as generic aspect — there are diverse concerns, e.g. pushing mobiles with specific coupons, that make use of it. Current range is mobile-app, browser, facebook-app, chat-app




There are also use cases for steering customer experience (incl. features and content) depending on device type. Via this header info should be passed-through as generic aspect. Current range is smartphone, tablet, desktop, other




On top of device type above, we even want to differ between device platform, e.g. smartphone Android vs. iOS. Via this header info should be passed-through as generic aspect. Current range is iOS, Android, Windows, Linux, MacOS




The app domain (i.e. shop channel context) of the request. Note, app-domain is a legacy concept that will be replaced in new platform by combinations of main CFA concerns like retailer, sales channel, country


Exception: The only exception to this guideline are the conventional hop-by-hop X-RateLimit- headers which can be used as defined in Must: Use 207 for Batch or Bulk Requests.

Must: Propagate Proprietary Headers

All Zalando’s proprietary headers are end-to-end headers. [2]

All headers specified above must be propagated to the services down the call chain. The header names and values must remain unchanged.

For example, the values of the custom headers like X-Device-Type can affect the results of queries by using device type information to influence recommendation results. Besides, the values of the custom headers can influence the results of the queries (e.g. the device type information influences the recommendation results).

Sometimes the value of a proprietary header will be used as part of the entity in a subsequent request. In such cases, the proprietary headers must still be propagated as headers with the subsequent request, despite the duplication of information.

16. Deprecation

Sometimes it is necessary to phase out an API endpoint (or version). I.e. this may be necessary if a field is no longer supported in the result or a whole business functionality behind an endpoint has to be shut down. There are many other reasons as well.

Must: Obtain Approval of Clients

Before shutting down an API (or version of an API) the producer must make sure, that all clients have given their consent to shut down the endpoint. Producers should help consumers to migrate to a potential new endpoint (i.e. by providing a migration manual). After all clients are migrated, the producer may shut down the deprecated API.

Must: External Partners Must Agree on Deprecation Timespan

If the API is consumed by any external partner, the producer must define a reasonable timespan that the API will be maintained after the producer has announced deprecation. The external partner (client) must agree to this minimum after-deprecation-lifespan before he starts using the API.

Must: Reflect Deprecation in API Definition

API deprecation must be part of the OpenAPI definition. If a method on a path, a whole path or even a whole API endpoint (multiple paths) should be deprecated, the producers must set deprecated=true on each method / path element that will be deprecated (OpenAPI 2.0 only allows you to define deprecation on this level). If deprecation should happen on a more fine grained level (i.e. query parameter, payload etc.), the producer should set deprecated=true on the affected method / path element and add further explanation to the description section.

If deprecated is set to true, the producer must describe what clients should use instead and when the API will be shut down in the description section of the API definition.

Must: Monitor Usage of Deprecated APIs

Owners of APIs used in production must monitor usage of deprecated APIs until the API can be shut down in order to align deprecation and avoid uncontrolled breaking effects. See also the Must: Provide Online Access to OpenAPI Reference Definition.

Should: Add a Warning Header to Responses

During deprecation phase, the producer should add a Warning header (see RFC 7234 - Warning header) field. When adding the Warning header, the warn-code must be 299 and the warn-text should be in form of "The path/operation/parameter/…​ {name} is deprecated and will be removed by {date}. Please see {link} for details." with a link to a documentation describing why the API is no longer supported in the current form and what clients should do about it. Adding the Warning header is not sufficient to gain client consent to shut down an API.

Should: Add Monitoring for Warning Header

Clients should monitor the Warning header in HTTP responses to see if an API will be deprecated in future.

Must: Not Start Using Deprecated APIs

Clients must not start using deprecated parts of an API.

17. API Operation

Must: Provide Online Access to OpenAPI Reference Definition

All service applications must support access to the OpenAPI Reference Definitions of their external APIs — it is optional for internal APIs — via the following two API endpoints:

  "schema_url": "/swagger.json",
  "schema_type": "swagger-2.0",
  "ui_url": "/ui/"
  • whether these endpoints have to be secured by OAuth depends on context of your API and product management

  • if you secure these endpoint, only use uid scope

Hint: Though discovery endpoints have to be supported, they should not be specified in the OpenAPI definition as they are generic and provide no API specific information.

We distinguish between internal and external APIs of an application which is owned by a specific team and often implemented via small set of services. An external API is used by clients outside the team - usually another application owned by a different team or even an external business partner user of our platform. An internal API is only used within the application and only by the owning team, for instance, for operational or implementation internal purposes.

Background: In our dynamic and complex service infrastructure, it is important to provide API client developers a central place with online access to the OpenAPI reference definitions of all running applications. As a part of the Zalando infrastructure, the .well-known/schema-discovery endpoint is used by the API Discovery to detect all API definitions. It checks all running applications via the endpoint above and stores the discovered API definitions. API Discovery itself provides a RESTful API as well as an API Viewer (Swagger-UI) for central access to all discovered API definitions.

Editorial: For the time being, this document is an appropriate place to mention this rule, even though it is not a RESTful API definition rule but related to service implementation obligations to support client developer API discovery.

Further reading:

Should: Monitor API Usage

Owners of APIs used in production should monitor API service to get information about its using clients. This information, for instance, is useful to identify potential review partner for API changes.

Hint: A preferred way of client detection implementation is by logging of the client-id retrieved from the OAuth token.

18. Events

Zalando’s architecture centers around decoupled microservices and in that context we favour asynchronous event driven approaches. The guidelines in this section focus on how to design and publish events intended to be shared for others to consume.

Events, Event Types and Categories.

Events are defined using an item called an Event Type. The Event Type allows events to have their structure declared with a schema by producers and understood by consumers. An Event Type declares standard information, such as a name, an owning application (and by implication, an owning team), a schema defining the event’s custom data, and a compatibility mode declaring how the schema will be evolved. Event Types also allow the declaration of validation and enrichment strategies for events, along with supplemental information such as how events can be partitioned in an event stream.

Event Types belong to a well known Event Category (such as a data change category), which provides extra information that is common to that kind of event.

Event Types can be published and made available as API resources for teams to use, typically in an Event Type Registry. Each event published can then be validated against the overall structure of its event type and the schema for its custom data.

The basic model described above was originally developed in the Nakadi project, which acts as a reference implementation of the event type registry, and as a validating publish/subscribe broker for event producers and consumers.

Must: Treat Events as part of the service interface

Events are part of a service’s interface to the outside world equivalent in standing to a service’s REST API. Services publishing data for integration must treat their events as a first class design concern, just as they would an API. For example this means approaching events with the "API first" principle in mind as described in the Introduction.

Must: Make Events available for review

Services publishing event data for use by others must make the event schema available for review.

Must: Ensure Event Type schemas conform to Open API’s Schema Object

Event type schema are defined in accordance with the Open API Schema Object specification, which uses a subset of JSON Schema Draft 4, and also adds other features. This allows events to properly align with API resource representations (it’s particulary useful for events that represent data changes about resources). The guideline exists since declaring event type schema using JSON-Schema syntax is common practice (because Open API doesn’t yet allow standalone object definitions).

In the rest of this section we’ll call out some of the more notable differences between Open API and JSON-Schema. Please note this is not a complete list - in general it’s recommended to familiarise yourself with Open API’s Schema Object (the details are available from the "Schema Object" section of that specification).

Open API removes some JSON-Schema keywords. These must not be used in event type schemas. The list of Open API object keywords can be seen here, but for convenience the list of non-available keywords relative to JSON-Schema are:

  • additionalItems

  • contains

  • patternProperties

  • dependencies

  • propertyNames

  • const

  • not

  • oneOf

Open API redefines some keywords:

  • additionalProperties: For event types that declare compatibility guarantees, there are recommended constraints around the use of this field. See the guideline "Avoid additionalProperties in event type definitions" for details.

Open API extends JSON-Schema with some keywords:

  • readOnly: events are logically immutable, so readOnly can be considered redundant, but harmless.

  • discriminator: discriminators exist to support polymorphism and act as an alternative to oneOf.

  • ^x-: patterned objects in the form of vendor extensions can be used in event type schema, but it might be the case that general purpose validators do not understand them to enforce a validation check, and fall back to must-ignore processing. A future version of the guidelines may define well known vendor extensions for events.

Must: Ensure that Events are registered as Event Types

In Zalando’s architecture, events are registered using a structure called an Event Type. The Event Type declares standard information as follows:

  • A well known event category, such as a general or data change category.

  • The name of the event type.

  • An owning application, and by implication, an owning team.

  • A schema defining the event payload.

  • The compatibility mode for the type.

Event Types allow easier discovery of event information and ensure that information is well-structured, consistent, and can be validated.

Event type owners must pay attention to the choice of compatibility mode. The mode provides a means to evolve thee schema. The range of modes are designed to be flexible enough so that producers can evolve schemas while not inadvertently breaking existing consumers:

  • none: Any schema modification is accepted, even if it might break existing producers or consumers. When validating events, undefined properties are accepted unless declared in the schema.

  • forward: A schema S1 is forward compatible if the previously registered schema, S0 can read events defined by S1 - that is, consumers can read events tagged with the latest schema version using the previous version as long as consumers follow the robustness principle described in the guideline’s API Design Principles.

  • compatible: This means changes are fully compatible. A new schema, S1, is fully compatible when every event published since the first schema version will validate against the latest schema. In compatible mode, only the addition of new optional properties and definitions to an existing schema is allowed. Other changes are forbidden.

The compatibility mode interact with revision numbers in the schema version field, which follows semantic versioning (MAJOR.MINOR.PATCH):

  • Changing an event type with compatibility mode compatible can lead to a PATCH or MINOR version revision. MAJOR breaking changes are not allowed.

  • Changing an event type with compatibility mode forward can lead to a PATCH or MINOR version revision. MAJOR breaking changes are not allowed.

  • Changing an event type with compatibility mode none can lead to PATCH, MINOR or MAJOR level changes.

The following examples illustrate this relations:

  • Changes to the event type’s title or description are considered PATCH level.

  • Adding new optional fields to an event type’s schema is considered a MINOR level change.

  • All other changes are considered MAJOR level, such as renaming or removing fields, or adding new required fields.

The core Event Type structure is shown below as an Open API object definition:

  description: |
    An event type defines the schema and its runtime properties. The required
    fields are the minimum set the creator of an event type is expected to
    - name
    - category
    - owning_application
    - schema
      description: |
        Name of this EventType.  Note: the name can encode the
        owner/responsible for this EventType and ideally should follow a
        naming pattern that makes it easy to read and understand.
      type: string
      pattern: '[a-zA-Z][-0-9a-zA-Z_]*(\.[a-zA-Z][-0-9a-zA-Z_]*)*'
      example: order.order_cancelled, business_partner.contract
      description: |
        Name of the application (eg, as would be used in infrastructure
        application or service registry) owning this `EventType`.
      type: string
      example: price-service
      description: Defines the category of this EventType.
      type: string
        - data
        - general
      description: |
        The compatibility mode to evolve the schema.
      type: string
        - compatible
        - forward
        - none
      default: forward
      description: The most recent payload schema for this EventType.
      type: object
          description: Values are based on semantic versioning (eg "1.2.1").
          type: string
          default: '1.0.0'
          description: Creation timestamp of the schema.
          type: string
          readOnly: true
          format: date-time
          example: '1996-12-19T16:39:57-08:00'
          description: |
             The schema language of schema definition. Currently only
             json_schema (JSON Schema v04) syntax is defined, but in the
             future there could be others.
          type: string
            - json_schema
          description: |
              The schema as string in the syntax defined in the field type.
          type: string
        - type
        - schema
      description: When this event type was created.
      type: string
      pattern: date-time
      description: When this event type was last updated.
      type: string
      pattern: date-time

APIs such as registries supporting event types, may extend the model, including the set of supported categories and schema formats. For example the Nakadi API’s event category registration also allows the declaration of validation and enrichment strategies for events, along with supplemental information, such as how events are partitioned in the stream.

Must: Ensure Events conform to a well-known Event Category

An event category describes a generic class of event types. The guidelines define two such categories:

  • General Event: a general purpose category.

  • Data Change Event: a category used for describing changes to data entities used for data replication based data integration.

The set of categories is expected to evolve in the future.

A category describes a predefined structure that event publishers must conform to along with standard information about that kind of event (such as the operation for a data change event).

The General Event Category.

The structure of the General Event Category is shown below as an Open API Schema Object definition:

  description: |
    A general kind of event. Event kinds based on this event define their
    custom schema payload as the top level of the document, with the
    "metadata" field being required and reserved for standard metadata. An
    instance of an event based on the event type thus conforms to both the
    EventMetadata definition and the custom schema definition. Previously
    this category was called the Business Category.
    - metadata
        $ref: '#/definitions/EventMetadata'

Event types based on the General Event Category define their custom schema payload at the top-level of the document, with the metadata field being reserved for standard information (the contents of metadata are described further down in this section).

In the example fragment below, the reserved metadata field is shown with fields "a" and "b" being defined as part of the custom schema:


  • The General Event in a previous version of the guidelines was called a Business Event. Implementation experience has shown that the category’s structure gets used for other kinds of events, hence the name has been generalized to reflect how teams are using it.

  • The General Event is still useful and recommended for the purpose of defining events that drive a business process.

  • The Nakadi broker still refers to the General Category as the Business Category and uses the keyword "business" for event type registration. Other than that, the JSON structures are identical.

See Must: Events must not provide sensitive customer personal data for more guidance on how to use the category.

The Data Change Event Category.

The Data Change Event Category structure is shown below as an Open API Schema Object:

  description: |
    Represents a change to an entity. The required fields are those
    expected to be sent by the producer, other fields may be added
    by intermediaries such as a publish/subscribe broker. An instance
    of an event based on the event type conforms to both the
    DataChangeEvent's definition and the custom schema definition.
    - metadata
    - data_op
    - data_type
    - data
      description: The metadata for this event.
      $ref: '#/definitions/EventMetadata'
      description: |
        Contains custom payload for the event type. The payload must conform
        to a schema associated with the event type declared in the metadata
        object's `event_type` field.
      type: object
      description: name of the (business) data entity that has been mutated
      type: string
      example: 'sales_order.order'
      type: string
      enum: ['C', 'U', 'D', 'S']
      description: |
        The type of operation executed on the entity:

        - C: Creation of an entity
        - U: An update to an entity.
        - D: Deletion of an entity.
        - S: A snapshot of an entity at a point in time.

The Data Change Event Category is structurally different to the General Event Category. It defines a field called data for placing the custom payload information, as well as specific information related to data changes in the data_type. In the example fragment below, the fields a and b are part of the custom payload housed inside the data field:

See the following guidelines for more guidance on how to use the Data Change Event Category:

Event Metadata.

The General and Data Change event categories share a common structure for metadata. The metadata structure is shown below as an Open API Schema Object:

  type: object
  description: |
    Carries metadata for an Event along with common fields. The required
    fields are those expected to be sent by the producer, other fields may be
    added by intermediaries such as publish/subscribe broker.
    - eid
    - occurred_at
      description: Identifier of this event.
      type: string
      format: uuid
      example: '105a76d8-db49-4144-ace7-e683e8f4ba46'
      description: The name of the EventType of this Event.
      type: string
      example: 'example.important-business-event'
      description: When the event was created according to the producer.
      type: string
      format: date-time
      example: '1996-12-19T16:39:57-08:00'
      description: |
        When the event was seen by an intermediary such as a broker.
      type: string
      readOnly: true
      format: date-time
      example: '1996-12-19T16:39:57-08:00'
      description: |
        Version of the schema used for validating this event. This may be
        enriched upon reception by intermediaries. This string uses semantic
      type: string
      readOnly: true
      description: |
        Event identifiers of the Event that caused the generation of
        this Event. Set by the producer.
      type: array
        type: string
        format: uuid
      example: '105a76d8-db49-4144-ace7-e683e8f4ba46'
      description: |
        A flow-id for this event (corresponds to the X-Flow-Id HTTP header).
      type: string
      example: 'JAh6xH4OQhCJ9PutIV_RYw'
      description: |
        Indicates the partition assigned to this Event. Used for systems
        where an event type's events can be sub-divided into partitions.
      type: string
      example: '0'

Please note than intermediaries acting between the producer of an event and its ultimate consumers, may perform operations like validation of events and enrichment of an event’s metadata. For example brokers such as Nakadi, can validate and enrich events with arbitrary additional fields that are not specified here and may set default or other values, if some of the specified fields are not supplied. How such systems work is outside the scope of these guidelines but producers and consumers working with such systems should be look into their documentation for additional information.

Must: Ensure that Events define useful business resources

Events are intended to be used by other services including business process/data analytics and monitoring. They should be based around the resources and business processes you have defined for your service domain and adhere to its natural lifecycle (see also Should: Model complete business processes).

As there is a cost in creating an explosion of event types and topics, prefer to define event types that are abstract/generic enough to be valuable for multiple use cases, and avoid publishing event types without a clear need.

Must: Events must not provide sensitive customer personal data

Similar to API permission scopes, there will be Event Type permissions passed via an OAuth token supported in near future. In the meantime, teams are asked to note the following:

  • Sensitive data, such as (e-mail addresses, phone numbers, etc) are subject to strict access and data protection controls.

  • Event type owners must not publish sensitive information unless it’s mandatory or necessary to do so. For example, events sometimes need to provide personal data, such as delivery addresses in shipment orders (as do other APIs), and this is fine.

Must: Use the General Event Category to signal steps and arrival points in business processes

When publishing events that represent steps in a business process, event types must be based on the General Event category.

All your events of a single business process will conform to the following rules:

  • Business events must contain a specific identifier field (a business process id or "bp-id") similar to flow-id to allow for efficient aggregation of all events in a business process execution.

  • Business events must contain a means to correctly order events in a business process execution. In distributed settings where monotonically increasing values (such as a high precision timestamp that is assured to move forwards) cannot be obtained, the parent_eids data structure allows causal relationships to be declared between events.

  • Business events should only contain information that is new to the business process execution at the specific step/arrival point.

  • Each business process sequence should be started by a business event containing all relevant context information.

  • Business events must be published reliably by the service.

At the moment we cannot state whether it’s best practice to publish all the events for a business process using a single event type and represent the specific steps with a state field, or whether to use multiple event types to represent each step. For now we suggest assessing each option and sticking to one for a given business process.

Must: Use Data Change Events to signal mutations

When publishing events that represents created, updated, or deleted data, change event types must be based on the Data Change Event category.

Should: Provide a means for explicit event ordering

Some common error cases may require event consumers to reconstruct event streams or replay events from a position within the stream. Events should therefore contain a way to restore their partial order of occurrence.

This can be done - among other ways - by adding - a strictly monotonically increasing entity version (e.g. as created by a database) to allow for partial ordering of all events for an entity - a strictly monotonically increasing message counter

System timestamps are not necessarily a good choice, since exact synchronization of clocks in distributed systems is difficult, two events may occur in the same microsecond and system clocks may jump backward or forward to compensate drifts or leap-seconds. If you use system timestamps to indicate event ordering, you must carefully ensure that your designated event order is not messed up by these effects.

Note that basing events on data structures that can be converged upon in a distributed setting (such as CRDTs, logical clocks and vector clocks) is outside the scope of this guidance.

Should: Use the hash partition strategy for Data Change Events

The hash partition strategy allows a producer to define which fields in an event are used as input to compute a logical partition the event should be added to. Partitions are useful as they allow supporting systems to scale their throughput while provide local ordering for event entities.

The hash option is particulary useful for data changes as it allows all related events for an entity to be consistently assigned to a partition, providing a relative ordered stream of events for that entity. This is because while each partition has a total ordering, ordering across partitions is not assured by a supporting system, thus it is possible for events sent across partitions to appear in a different order to consumers that the order they arrived at the server.

When using the hash strategy the partition key in almost all cases should represent the entity being changed and not a per event or change identifier such as the eid field or a timestamp. This ensures data changes arrive at the same partition for a given entity and can be consumed effectively by clients.

There may be exceptional cases where data change events could have their partition strategy set to be the producer defined or random options, but generally hash is the right option - that is while the guidelines here are a "should", they can be read as "must, unless you have a very good reason".

Should: Ensure that Data Change Events match API representations

A data change event’s representation of an entity should correspond to the REST API representation.

There’s value in having the fewest number of published structures for a service. Consumers of the service will be working with fewer representations, and the service owners will have less API surface to maintain. In particular, you should only publish events that are interesting in the domain and abstract away from implementation or local details - there’s no need to reflect every change that happens within your system.

There are cases where it could make sense to define data change events that don’t directly correspond to your API resource representations. Some examples are -

  • Where the API resource representations are very different from the datastore representation, but the physical data are easier to reliably process for data integration.

  • Publishing aggregated data. For example a data change to an individual entity might cause an event to be published that contains a coarser representation than that defined for an API

  • Events that are the result of a computation, such as a matching algorithm, or the generation of enriched data, and which might not be stored as entity by the service.

Must: Permissions on events must correspond to API permissions

If a resource can be read synchronously via a REST API and read asynchronously via an event, the same read-permission must apply: We want to protect access to data, not the way data is accessed.

Must: Indicate ownership of Event Types

Event definitions must have clear ownership - this can be indicated via the owning_application field of the EventType.

Typically there is one producer application, which owns the EventType and is responsible for its definition, akin to how RESTful API definitions are managed. However, the owner may also be a particular service from a set of multiple services that are producing the same kind of event.

Must: Define Event Payloads in accordance with the overall Guidelines

Events must be consistent with other API data and the API Guidelines in general.

Everything expressed in the Introduction to these Guidelines is applicable to event data interchange between services. This is because our events, just like our APIs, represent a commitment to express what our systems do and designing high-quality, useful events allows us to develop new and interesting products and services.

What distinguishes events from other kinds of data is the delivery style used, asynchronous publish-subscribe messaging. But there is no reason why they could not be made available using a REST API, for example via a search request or as a paginated feed, and it will be common to base events on the models created for the service’s REST API.

The following existing guideline sections are applicable to events:

Must: Maintain backwards compatibility for Events

Changes to events must be based around making additive and backward compatible changes. This follows the guideline, "Must: Don’t Break Backward Compatibility" from the Compatibility guidelines.

In the context of events, compatibility issues are complicated by the fact that producers and consumers of events are highly asynchronous and can’t use content-negotiation techniques that are available to REST style clients and servers. This places a higher bar on producers to maintain compatibility as they will not be in a position to serve versioned media types on demand.

For event schema, these are considered backward compatible changes, as seen by consumers -

  • Adding new optional fields to JSON objects.

  • Changing the order of fields (field order in objects is arbitrary).

  • Changing the order of values with same type in an array.

  • Removing optional fields.

  • Removing an individual value from an enumeration.

These are considered backwards-incompatible changes, as seen by consumers -

  • Removing required fields from JSON objects.

  • Changing the default value of a field.

  • Changing the type of a field, object, enum or array.

  • Changing the order of values with different type in an array (also known as a tuple).

  • Adding a new optional field to redefine the meaning of an existing field (also known as a co-occurrence constraint).

  • Adding a value to an enumeration (note that x-extensible-enum is not available in JSON Schema)

Should: Avoid additionalProperties in event type definitions

Event type schema should avoid using additionalProperties declarations, in order to support schema evolution.

Events are often intermediated by publish/subscribe systems and are commonly captured in logs or long term storage to be read later. In particular, the schemas used by publishers and consumers can
drift over time. As a result, compatibility and extensibility issues that happen less frequently with client-server style APIs become important and regular considerations for event design. The guidelines recommend the following to enable event schema evolution:

  • Publishers who intend to provide compatibility and allow their schemas to evolve safely over time must not declare an additionalProperties field with a value of true (i.e., a wildcard extension point). Instead they must define new optional fields and update their schemas in advance of publishing those fields.

  • Consumers must ignore fields they cannot process and not raise errors. This can happen if they are processing events with an older copy of the event schema than the one containing the new definitions specified by the publishers.

The above constraint does not mean fields can never be added in future revisions of an event type schema - additive compatible changes are allowed, only that the new schema for an event type must define the field first before it is published within an event. By the same turn the consumer must ignore fields it does not know about from its copy of the schema, just as they would as an API client - that is, they cannot treat the absence of an additionalProperties field as though the event type schema was closed for extension.

Requiring event publishers to define their fields ahead of publishing avoids the problem of field redefinition. This is when a publisher defines a field to be of a different type that was already being emitted, or, is changing the type of an undefined field. Both of these are prevented by not using additionalProperties.

See also "Treat Open API Definitions As Open For Extension By Default"
in the Compatibility section for further guidelines on the use of additionalProperties.

Must: Use unique Event identifiers

The eid (event identifier) value of an event must be unique.

The eid property is part of the standard metadata for an event and gives the event an identifier. Producing clients must generate this value when sending an event and it must be guaranteed to be unique from the perspective of the owning application. In particular events within a given event type’s stream must have unique identifiers. This allows consumers to process the eid to assert the event is unique and use it as an idempotency check.

Note that uniqueness checking of the eid might be not enforced by systems consuming events and it is the responsibility of the producer to ensure event identifiers do in fact distinctly identify events. A straightforward way to create a unique identifier for an event is to generate a UUID value.

Should: Design for idempotent out-of-order processing

Events that are designed for idempotent out-of-order processing allow for extremely resilient systems: If processing an event fails, consumers and producers can skip/delay/retry it without stopping the world or corrupting the processing result.

To enable this freedom of processing, you must explicitly design for idempotent out-of-order processing: Either your events must contain enough information to infer their original order during consumption or your domain must be designed in a way that order becomes irrelevant.

As common example similar to data change events, idempotent out-of-order processing can be supported by sending the following information:

A receiver that is interested in the current state can then ignore events that are older than the last processed event of each resource. A receiver interested in the history of a resource can use the ordering key to recreate a (partially) ordered sequence of events.

Must: Follow conventions for Event Type names

Event types can follow these naming conventions (each convention has its own should, must or could conformance level) -

  • Event type names must be url-safe. This is because the event type names may appear in URLs published by other systems and APIs.

  • Event type names should be lowercase words and numbers, using hyphens, underscores or periods as separators.

Must: Prepare for duplicate Events

Event consumers must be able to process duplicate events.

Most message brokers and data streaming systems offer “at-least-once” delivery. That is, one particular event is delivered to the consumers one or more times. Other circumstances can also cause duplicate events.

For example, these situations occur if the publisher sends an event and doesn’t receive the acknowledgment (e.g. due to a network issue). In this case, the publisher will try to send the same event again. This leads to two identical events in the event bus which have to be processed by the consumers. Similar conditions can appear on consumer side: an event has been processed successfully, but the consumer fails to confirm the processing.

Appendix A: References

Appendix B: Tooling

This is not a part of the actual guidelines, but might be helpful for following them. Using a tool mentioned here doesn’t automatically ensure you follow the guidelines.

Appendix C: Changelog

This change log only contains major changes made after October 2016.

Non-major changes are editorial-only changes or minor changes of existing guidelines, e.g. adding new error code. Major changes are changes that come with additional obligations, or even change an existing guideline obligation. The latter changes are additionally labeled with "Rule Change" here.

To see a list of all changes, please have a look at the commit list in Github.

Rule Changes

  • 2017-08-22: Migration to Asciidoc

  • 2017-07-20: Be more precise on client vs. server obligations for compatible API extensions.

  • 2017-06-06: Made money object guideline clearer.

  • 2017-05-17: Added guideline on query parameter collection format.

  • 2017-05-10: Added the convention of using RFC2119 to describe guideline levels, and replaced book.could with book.may.

  • 2017-03-30: Added rule that permissions on resources in events must correspond to permissions on API resources

  • 2017-03-30: Added rule that APIs should be modelled around business processes

  • 2017-02-28: Extended information about how to reference sub-resources and the usage of composite identifiers in the Must: Use Domain-Specific Resource Names part.

  • 2017-02-22: Added guidance for conditional requests with If-Match/If-None-Match

  • 2017-02-02: Added guideline for batch and bulk request

  • 2017-02-01: May: Use Content-Location Header

  • 2017-01-18: Removed "Avoid Javascript Keywords" rule

  • 2017-01-05: Clarification on the usage of the term "REST/RESTful"

  • 2016-12-07: Introduced "API as a Product" principle

  • 2016-12-06: New guideline: "Should Only Use UUIDs If Necessary"

  • 2016-12-04: Changed OAuth flow example from implicit to password in Security.

  • 2016-10-13: Must: Define Format for Type Number and Integer

  • 2016-10-10: Introduced the changelog. From now on all rule changes on API guidelines will be recorded here.

1. Per definition of R.Fielding REST APIs have to support HATEOAS (maturity level 3). Our guidelines do not strongly advocate for full REST compliance, but limited hypermedia usage, e.g. for pagination (see Hypermedia). However, we still use the term "RESTful API", due to the absence of an alternative established term and to keep it like the very majority of web service industry that also use the term for their REST approximations — in fact, in today’s industry full HATEOAS compliant APIs are a very rare exception.
2. HTTP/1.1 standard (RFC-7230) defines two types of headers: end-to-end and hop-by-hop headers. End-to-end headers must be transmitted to the ultimate recipient of a request or response. Hop-by-hop headers, on the contrary, are meaningful for a single connection only.