RFC 8977 | RDAP Sorting and Paging | January 2021 |
Loffredo, et al. | Standards Track | [Page] |
The Registration Data Access Protocol (RDAP) does not include core functionality for clients to provide sorting and paging parameters for control of large result sets. This omission can lead to unpredictable server processing of queries and client processing of responses. This unpredictability can be greatly reduced if clients can provide servers with their preferences for managing large responses. This document describes RDAP query extensions that allow clients to specify their preferences for sorting and paging result sets.¶
This is an Internet Standards Track document.¶
This document is a product of the Internet Engineering Task Force (IETF). It represents the consensus of the IETF community. It has received public review and has been approved for publication by the Internet Engineering Steering Group (IESG). Further information on Internet Standards is available in Section 2 of RFC 7841.¶
Information about the current status of this document, any errata, and how to provide feedback on it may be obtained at https://www.rfc-editor.org/info/rfc8977.¶
Copyright (c) 2021 IETF Trust and the persons identified as the document authors. All rights reserved.¶
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.¶
The availability of functionality for result sorting and paging provides benefits to both clients and servers in the implementation of RESTful services [REST]. These benefits include:¶
Approaches to implementing features for result sorting and paging can be grouped into two main categories:¶
Sorting and paging are implemented through the introduction of additional parameters in the query string (e.g., the Open Data Protocol (ODATA) [ODATA-PART1]).¶
However, there are some drawbacks associated with the use of the HTTP header. First, the header properties cannot be set directly from a web browser. Moreover, in an HTTP session, the information on the status (i.e., the session identifier) is usually inserted in the header or a cookie, while the information on the resource identification or the search type is included in the query string. Finally, providing custom information through HTTP headers assumes the client has prior knowledge of the server implementation, which is widely considered a Representational State Transfer (REST) design anti-pattern. As a result, this document describes a specification based on the use of query parameters.¶
Currently, RDAP [RFC7482] defines two query types:¶
While the lookup query does not raise issues regarding response size management, the search query can potentially generate a large result set that is often truncated according to server limits. Besides, it is not possible to obtain the total number of objects found that might be returned in a search query response [RFC7483]. Lastly, there is no way to specify sort criteria to return the most relevant objects at the beginning of the result set. Therefore, the client might traverse the whole result set to find the relevant objects or, due to truncation, might not find them at all.¶
The specification described in this document extends RDAP query capabilities to enable result sorting and paging by adding new query parameters that can be applied to RDAP search path segments. The service is implemented using the Hypertext Transfer Protocol (HTTP) [RFC7230] and the conventions described in [RFC7480].¶
The implementation of the new parameters is technically feasible, as operators for counting, sorting, and paging rows are currently supported by the major relational database management systems.¶
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶
The new query parameters are OPTIONAL extensions of path segments defined in [RFC7482]. They are as follows:¶
Augmented Backus-Naur Form (ABNF) [RFC5234] is used in the following sections to describe the formal syntax of these new parameters.¶
According to most advanced principles in REST design, collectively known as HATEOAS (Hypermedia as the Engine of Application State) [HATEOAS], a client entering a REST application through an initial URI should use server-provided links to dynamically discover available actions and access the resources it needs. In this way, the client is neither required to have prior knowledge of the service nor, consequently, to hard code the URIs of different resources. This allows the server to make URI changes as the API evolves without breaking clients. Definitively, a REST service should be as self-descriptive as possible.¶
Therefore, servers implementing the query parameters described in this specification SHOULD provide additional information in their responses about both the available sorting criteria and possible pagination. Such information is collected in two OPTIONAL response elements named "sorting_metadata" and "paging_metadata".¶
The "sorting_metadata" element contains the following properties:¶
An array of objects, with each element describing an available sort criterion. The AvailableSort object includes the following members:¶
At least one of the "currentSort" and "availableSorts" properties MUST be present.¶
The "paging_metadata" element contains the following fields:¶
Servers returning the "paging_metadata" element in their response MUST include the string literal "paging" in the rdapConformance array. Servers returning the "sorting_metadata" element MUST include the string literal "sorting".¶
Currently, RDAP does not allow a client to determine the total number of results in a query response when the result set is truncated. This is inefficient because the user cannot determine if the result set is complete.¶
The "count" parameter provides additional functionality that allows a client to request information from the server that specifies the total number of objects matching the search pattern.¶
The following is an example of an RDAP query including the "count" parameter:¶
https://example.com/rdap/domains?name=example*.com&count=true¶
The ABNF syntax is the following:¶
count = "count=" ( trueValue / falseValue ) trueValue = ("true" / "yes" / "1") falseValue = ("false" / "no" / "0")¶
A trueValue means that the server MUST provide the total number of objects in the "totalCount" field of the "paging_metadata" element (Figure 1). A falseValue means that the server MUST NOT provide this number.¶
RDAP does not provide any capability to specify the result set sort criteria. A server could implement a default sorting scheme according to the object class, but this feature is not mandatory and might not meet user requirements. Sorting can be addressed by the client, but this solution is rather inefficient. Sorting features provided by the RDAP server could help avoid truncation of relevant results.¶
The "sort" parameter allows the client to ask the server to sort the results according to the values of one or more properties and according to the sort direction of each property. The ABNF syntax is the following:¶
sort = "sort=" sortItem *( "," sortItem ) sortItem = property-ref [":" ( "a" / "d" ) ] property-ref = ALPHA *( ALPHA / DIGIT / "_" )¶
"a" means that an ascending sort MUST be applied; "d" means that a descending sort MUST be applied. If the sort direction is absent, an ascending sort MUST be applied.¶
The following are examples of RDAP queries that include the "sort" parameter:¶
https://example.com/rdap/domains?name=example*.com&sort=name¶
https://example.com/rdap/ domains?name=example*.com&sort=registrationDate:d¶
https://example.com/rdap/ domains?name=example*.com&sort=lockedDate,name¶
Except for sorting IP addresses and values denoting dates and times, servers MUST implement sorting according to the JSON value type of the RDAP field the sorting property refers to. That is, JSON strings MUST be sorted lexicographically, and JSON numbers MUST be sorted numerically. Values denoting dates and times MUST be sorted in chronological order. If IP addresses are represented as JSON strings, they MUST be sorted based on their numeric conversion.¶
The conversion of an IPv4 address to a number is possible since each dotted format IPv4 address is a representation of a number written in a 256-based manner; for example, 192.168.0.1 means 1*256^0 + 0*256^1 + 168*256^2 + 192*256^3 = 3232235521. Similarly, an IPv6 address can be converted into a number by applying the base 65536. Therefore, the numerical representation of the IPv6 address 2001:0db8:85a3:0:0:8a2e:0370:7334 is 42540766452641154071740215577757643572. Built-in functions and libraries for converting IP addresses into numbers are available in most known programming languages and relational database management systems.¶
If the "sort" parameter presents an allowed sorting property, it MUST be provided in the "currentSort" field of the "sorting_metadata" element.¶
In the "sort" parameter ABNF syntax, the element named "property-ref" represents a reference to a property of an RDAP object. Such a reference could be expressed by using a JSONPath expression (named "jsonpath" in the following).¶
JSONPath is a syntax, originally based on the XML XPath notation [W3C.CR-xpath-31-20170321], which represents a path to select an element (or a set of elements) in a JSON document [RFC8259]. For example, the jsonpath to select the value of the ASCII name inside an RDAP domain lookup response is "$.ldhName", where $ identifies the root of the document object model (DOM). Another way to select a value inside a JSON document is the JSON Pointer [RFC6901].¶
While JSONPath and JSON Pointer are both commonly adopted notations to select any value inside JSON data, neither is particularly concise and easy to use (e.g., "$.domainSearchResults[*].events[?(@.eventAction='registration')].eventDate" is the jsonpath of the registration date in an RDAP domain search response).¶
Therefore, this specification defines the "property-ref" element in terms of names identifying RDAP properties. However, not all the RDAP properties are suitable to be used in sort criteria. These properties include:¶
properties providing service information (e.g., links, notices, and remarks)¶
multivalued properties (e.g., status, roles, and variants)¶
On the contrary, properties expressed as values of other properties (e.g., registration date) could be used in such a context.¶
A list of properties an RDAP server MAY implement is defined. The properties are divided into two groups: object-common properties and object-specific properties.¶
Object-common properties. Object-common properties are derived from merging the "eventAction" and the "eventDate" properties. The following values of the "sort" parameter are defined:¶
Object-specific properties. Note that some of these properties are also defined as query path segments. These properties include:¶
The correspondence between these sorting properties and the RDAP object classes is shown in Table 1. Some of the sorting properties defined for the RDAP entity class are related to jCard elements [RFC7095], but because jCard is the JSON format for vCard, the corresponding definitions are included in the vCard specification [RFC6350].¶
An RDAP server MUST NOT use the defined sorting properties with a meaning other than that described in Table 1.¶
Object class | Sorting property | RDAP property | RFC 7483 | RFC 6350 | RFC 8605 |
---|---|---|---|---|---|
Searchable objects | Common properties | eventAction values suffixed by "Date" | 4.5 | ||
Domain | name | unicodeName/ ldhName | 5.3 | ||
Nameserver | name | unicodeName/ ldhName | 5.2 | ||
ipv4 | v4 ipAddress | 5.2 | |||
ipv6 | v6 ipAddress | 5.2 | |||
Entity | handle | handle | 5.1 | ||
fn | jCard fn | 5.1 | 6.2.1 | ||
org | jCard org | 5.1 | 6.6.4 | ||
voice | jCard tel with type="voice" | 5.1 | 6.4.1 | ||
jCard email | 5.1 | 6.4.2 | |||
country | country name in jCard adr | 5.1 | 6.3.1 | ||
cc | country code in jCard adr | 5.1 | 3.1 | ||
city | locality in jCard adr | 5.1 | 6.3.1 |
Regarding the definitions in Table 1, some further considerations are needed to disambiguate some cases:¶
Since the response to a search on either domains or nameservers might include both A-labels and U-labels [RFC5890] in general, a consistent sorting policy MUST treat the unicodeName and ldhName as two representations of the same value. The unicodeName value MUST be used while sorting if it is present; when the unicodeName is unavailable, the value of the ldhName MUST be used instead.¶
The jCard "sort-as" parameter MUST be ignored for the sorting capability described in this document.¶
Even if a nameserver can have multiple IPv4 and IPv6 addresses, the most common configuration includes one address for each IP version. Therefore, this specification makes the assumption that nameservers have a single IPv4 and/or IPv6 value. When more than one address per IP version is presented, sorting MUST be applied to the first value.¶
Multiple events with a given action on an object might be returned. If this occurs, sorting MUST be applied to the most recent event.¶
The "jsonPath" field in the "sorting_metadata" element is used to clarify the RDAP response field the sorting property refers to. The mapping between the sorting properties and the jsonpaths of the RDAP response fields is shown below. The JSONPath operators used herein are described in Appendix A.¶
Searchable objects¶
$.domainSearchResults[*].events[?(@.eventAction=="registration")].eventDate¶
$.domainSearchResults[*].events[?(@.eventAction=="reregistration")].eventDate¶
$.domainSearchResults[*].events[?(@.eventAction=="last changed")].eventDate¶
$.domainSearchResults[*].events[?(@.eventAction=="expiration")].eventDate¶
$.domainSearchResults[*].events[?(@.eventAction=="deletion")].eventDate¶
$.domainSearchResults[*].events[?(@.eventAction=="reinstantiation")].eventDate¶
$.domainSearchResults[*].events[?(@.eventAction=="transfer")].eventDate¶
$.domainSearchResults[*].events[?(@.eventAction=="locked")].eventDate¶
$.domainSearchResults[*].events[?(@.eventAction=="unlocked")].eventDate¶
Domain¶
$.domainSearchResults[*].[unicodeName,ldhName]¶
Nameserver¶
Entity¶
$.entitySearchResults[*].handle¶
$.entitySearchResults[*].vcardArray[1][?(@[0]=="fn")][3]¶
$.entitySearchResults[*].vcardArray[1][?(@[0]=="org")][3]¶
$.entitySearchResults[*].vcardArray[1][?(@[0]=="tel" && @[1].type=="voice")][3]¶
$.entitySearchResults[*].vcardArray[1][?(@[0]=="email")][3]¶
$.entitySearchResults[*].vcardArray[1][?(@[0]=="adr")][3][6]¶
$.entitySearchResults[*].vcardArray[1][?(@[0]=="adr")][1].cc¶
$.entitySearchResults[*].vcardArray[1][?(@[0]=="adr")][3][3]¶
Additional notes on the provided jsonpaths:¶
Those related to the event dates are defined only for the "domain" object. To obtain the equivalent jsonpaths for "entity" and "nameserver", the path segment "domainSearchResults" must be replaced with "entitySearchResults" and "nameserverSearchResults", respectively.¶
An RDAP server MAY use the "links" array of the "sorting_metadata" element to provide ready-made references [RFC8288] to the available sort criteria (Figure 2). Each link represents a reference to an alternate view of the results.¶
The "value", "rel", and "href" JSON values MUST be specified. All other JSON values are OPTIONAL.¶
The "cursor" parameter defined in this specification can be used to encode information about any pagination method. For example, in the case of a simple implementation of the "cursor" parameter to represent offset pagination information, the "cursor" value "b2Zmc2V0PTEwMCxsaW1pdD01MA==" is the base64 encoding of "offset=100,limit=50". Likewise, in a simple implementation to represent keyset pagination information, the "cursor" value "ZXhhbXBsZS1OLmNvbQ==" represents the base64 encoding of "key=example-N.com" whereby the key value identifies the last row of the current page.¶
Note that this specification uses a base64 encoding for cursor obfuscation just for example. RDAP servers are NOT RECOMMENDED to obfuscate the "cursor" value through a mere base64 encoding.¶
This solution lets RDAP providers implement a pagination method according to their needs, a user's access level, and the submitted query. Besides, servers can change the method over time without announcing anything to clients. The considerations that have led to this solution are described in more detail in Appendix B.¶
The ABNF syntax of the "cursor" parameter is the following:¶
cursor = "cursor=" 1*( ALPHA / DIGIT / "/" / "=" / "-" / "_" )¶
The following is an example of an RDAP query including the "cursor" parameter:¶
https://example.com/rdap/domains?name=example*.com &cursor=wJlCDLIl6KTWypN7T6vc6nWEmEYe99Hjf1XY1xmqV-M=¶
An RDAP server SHOULD use the "links" array of the "paging_metadata" element to provide a ready-made reference [RFC8288] to the next page of the result set (Figure 3). Examples of additional "rel" values a server MAY implement are "first", "last", and "prev".¶
The constraints for the values of parameters are defined by their ABNF syntax. Therefore, each request that includes an invalid value for a parameter SHOULD produce an HTTP 400 (Bad Request) response code. The same response SHOULD be returned in the following cases:¶
if sorting by either single or multiple properties, the client provides an unsupported value for the "sort" parameter, as well as a value related to an object property not included in the response¶
Optionally, the response MAY include additional information regarding either the supported sorting properties or the correct "cursor" value in the HTTP entity body (Figure 4).¶
Implementation of the new parameters is technically feasible, as operators for counting, sorting, and paging are currently supported by the major relational database management systems. Similar operators are completely or partially supported by the most well-known NoSQL databases (e.g., MongoDB, CouchDB, HBase, Cassandra, Hadoop, etc.). Additional implementation notes are included in Appendix C.¶
IANA has registered the following values in the "RDAP Extensions" registry:¶
Security services for the operations specified in this document are described in [RFC7481].¶
A search query typically requires more server resources (such as memory, CPU cycles, and network bandwidth) when compared to a lookup query. This increases the risk of server resource exhaustion and subsequent denial of service. This risk can be mitigated by either restricting search functionality or limiting the rate of search requests. Servers can also reduce their load by truncating the results in a response. However, this last security policy can result in a higher inefficiency or risk due to acting on incomplete information if the RDAP server does not provide any functionality to return the truncated results.¶
The new parameters presented in this document provide RDAP operators with a way to implement a server that reduces inefficiency risks. The "count" parameter gives the client the ability to evaluate the completeness of a response. The "sort" parameter allows the client to obtain the most relevant information at the beginning of the result set. This can reduce the number of unnecessary search requests. Finally, the "cursor" parameter enables the user to scroll the result set by submitting a sequence of sustainable queries within server-acceptable limits.¶
The jsonpaths used in this document are provided according to the Goessner proposal [GOESSNER-JSON-PATH].¶
Such specification requires that implementations support a set of "basic operators". These operators are used to access the elements of a JSON structure like objects and arrays, as well as their subelements (object members and array items, respectively). No operations are defined for retrieving parent or sibling elements of a given element. The root element is always referred to as $ regardless of it being an object or array.¶
Additionally, the specification permits implementations to support arbitrary script expressions. These can be used to index into an object or array, or to filter elements from an array. While script expression behavior is implementation-defined, most implementations support the basic relational and logical operators as well as both object member and array item access, sufficiently similar for the purpose of this document. Commonly supported operators/functions divided into "top-level operators" and "filter operators" are documented in Tables 2 and 3, respectively.¶
For more information on implementation interoperability issues, see [JSONPATH-COMPARISON]. At the time of writing, work is beginning on a standardization effort too (see [JSONPATH-WG]).¶
Operator | Description |
---|---|
$ | Root element |
.<name> | Object member access (dot-notation) |
['<name>'] | Object member access (bracket-notation) |
[<number>] | Array item access |
* | All elements within the specified scope |
[?(<expression>)] | Filter expression |
Operator | Description |
---|---|
@ | Current element being processed |
.<name> | Object member access |
.[<name1>,<name2>] | Union of object members |
[<number>] | Array item access |
== | Left is equal to right |
!= | Left is not equal to right |
< | Left is less than right |
<= | Left is less than or equal to right |
> | Left is greater than right |
>= | Left is greater than or equal to right |
&& | Logical conjunction |
|| | Logical disjunction |
An RDAP query could return a response with hundreds, even thousands, of objects, especially when partial matching is used. For this reason, the "cursor" parameter addressing result pagination is defined to make responses easier to handle.¶
Presently, the most popular methods to implement pagination in a REST API include offset pagination and keyset pagination. Neither pagination method requires the server to handle the result set in a storage area across multiple requests since a new result set is generated each time a request is submitted. Therefore, they are preferred to any other method requiring the management of a REST session.¶
Using limit and offset operators represents the traditionally used method to implement result pagination. Both of them can be used individually:¶
When limit and offset are used together, they provide the ability to identify a specific portion of the result set. For example, the pair "offset=100,limit=50" returns the first 50 objects starting from position 101 of the result set.¶
Though easy to implement, offset pagination also includes drawbacks:¶
When offset has a very high value, scrolling the result set could take some time.¶
It always requires fetching all rows before dropping as many rows as specified by offset.¶
Keyset pagination [SEEK] adds a query condition that enables the selection of the only data not yet returned. This method has been taken as the basis for the implementation of a "cursor" parameter [CURSOR] by some REST API providers [CURSOR-API1] [CURSOR-API2]. The cursor is a URL-safe string opaque to the client and representing a logical pointer to the first result of the next page.¶
Nevertheless, even keyset pagination can be troublesome:¶
It needs at least one key field.¶
It does not allow sorting simply by any field because the sorting criterion must contain a key.¶
It works best with full composite values supported by database management systems (i.e., [x,y]>[a,b]); emulation is possible but inelegant and less efficient.¶
It does not allow direct navigation to arbitrary pages because the result set must be scrolled in sequential order starting from the initial page.¶
Some additional considerations can be made in the RDAP context:¶
An RDAP object is a conceptual aggregation of information generally collected from more than one data structure (e.g., table), and this makes it even harder to implement keyset pagination, a task that is already quite difficult. For example, the entity object can include information from different data structures (registrars, registrants, contacts, resellers), each one with its key field mapping the RDAP entity handle.¶
Depending on the number of page results as well as the number and the complexity of the properties of each RDAP object in the response, the time required by offset pagination to skip the previous pages could be much faster than the processing time needed to build the current page. In fact, RDAP objects are usually formed by information belonging to multiple data structures and containing multivalued properties (i.e., arrays); therefore, data selection might be a time-consuming process. This situation occurs even though the selection is supported by indexes.¶
Ultimately, both pagination methods have benefits and drawbacks.¶
This section contains an overview of the main choices made during the implementation of the capabilities defined in this document in the RDAP public test server of Registro.it at the Institute of Informatics and Telematics of the National Research Council (IIT-CNR). The content of this section can represent guidance for implementers who plan to provide RDAP users with those capabilities. The RDAP public test server can be accessed at <https://rdap.pubtest.nic.it/>. Further documentation about the server features is available at <https://rdap.pubtest.nic.it/doc/README.html>.¶
If no sort criterion is specified in the query string, the results are sorted by a default property: "name" for domains and nameservers, and "handle" for entities. The server supports multiple property sorting but the "sorting_metadata" object includes only the links to alternative result set views sorted by a single property just to show the list of sorting properties allowed for each searchable object. The server supports all the object-specific sorting properties described in the specification except for nameserver sorting based on unicodeName, that is, the "name" sorting property is mapped onto the "ldhName" response field. Regarding the object-common properties, sorting by registrationDate, expirationDate, lastChangedDate, and transferDate is supported.¶
The counting operation is implemented through a separate query. Some relational database management systems support custom operators to get the total count together with the rows, but the resulting query can be considerably more expensive than that performed without the total count. Therefore, as "totalCount" is an optional response information, always fetching the total number of rows has been considered an inefficient solution. Furthermore, to avoid the processing of unnecessary queries, when the "count" parameter is included in the submitted query, it is not also repeated in the query strings of the "links" array provided in both "paging_metadata" and "sorting_metadata" objects.¶
The server implements the cursor pagination through the keyset pagination when sorting by a unique property is requested or the default sort is applied. Otherwise, it implements the cursor pagination through the offset pagination. As most relational database management systems don't support the comparison of full composite values natively, the implementation of full keyset pagination seem to be troublesome so, at least initially, a selective applicability of keyset pagination is advisable. Moreover, the "cursor" value encodes not only information about pagination but also about the search pattern and the other query parameters in order to check the consistency of the entire query string. If the "cursor" value is inconsistent with the rest of the query string, the server returns an error response.¶
The authors would like to acknowledge Brian Mountford, Tom Harrison, Karl Heinz Wolf, Jasdip Singh, Erik Kline, Éric Vyncke, Benjamin Kaduk, and Roman Danyliw for their contributions to the development of this document.¶