This document makes several assumptions:

• The type of client I am using is of interest to the people in my roster.
• Clients for the people on my roster might want to make user interface decisions based on my capabilities.
• Different clients may support the same capabilities.
• Members of a community tend to cluster around a small set of clients with a small set of capabilities. More specifically, multiple people in my roster use the same client, and they upgrade versions relatively slowly (commonly a few times a year, perhaps once a week at most, certainly not once a minute).
• Some clients are running on networks without server-to-server connectivity enabled and without access to the Internet via HTTP.
• Conversations are possible between users who are not on each other's rosters.
• Client capabilities may change over the course of a presence session, as features are enabled or disabled.

The protocol defined herein addresses the following requirements:

1. Clients must be able to participate even if they support only &xmppcore;, &xmppim;, and XEP-0030.
2. Clients must be able to participate even if they are on networks without connectivity to other XMPP servers, services offering specialized XMPP extensions, or HTTP servers.These first two requirements effectively eliminated &xep0060; as a possible implementation of entity capabilities.
3. Clients must be able to retrieve information without querying every entity with which they communicate.
4. Since presence is normally broadcasted to many contacts, the byte size of the proposed extension must be as small as possible.
5. It must be possible to write a XEP-0045 server implementation that passes the given information along.
6. It must be possible to publish a change in capabilities within a single presence session.
7. Server infrastructure above and beyond that defined in XMPP Core and XMPP IM must not be required for this approach to work, although additional server infrastructure may be used for optimization purposes.
8. The defined mechanism must not be limited to clients but must be usable be servers, components, and other network entities.

Entity capabilities are encapsulated in a <c/> element qualified by the 'http://jabber.org/protocol/caps' namespace. The attributes of the <c/> element are as follows.

* Note: It is RECOMMENDED for the value of the 'node' attribute to be an HTTP URL at which a user could find further information about the software product, such as "http://psi-im.org/" for the Psi client; this enables a processing application to also determine a unique string for the generating application, which it could maintain in a list of known software implementations (e.g., associating the name received via the disco#info reply with the URL found in the caps data).

** Note: Before version 1.5 of this specification, the version information was contained in the 'ver' attribute as described below.

*** Note: Before version 1.4 of this specification, the 'ver' attribute was used to specify the released version of the software; while the values of the 'ver' attribute that result from use of the algorithm specified herein are backwards-compatible, applications SHOULD appropriately handle the Legacy Format.

In order to help prevent poisoning of entity capabilities information, the value of the 'ver' attribute MUST be generated according to the following method.

Note: All sorting operations MUST be performed using "i;octet" collation as specified in Section 9.3 of &rfc4790;.

1. Initialize an empty string S.
2. Sort the service discovery identities A registry of service discovery identities is located at &DISCOCATEGORIES;. by category and then by type (if it exists), formatted as 'category' '/' 'type'.
3. For each identity, append the 'category/type' to S, followed by the '<' character.
4. Sort the supported service discovery features. A registry of service discovery features is located at &DISCOFEATURES;.
5. For each feature, append the feature to S, followed by the '<' character.
6. Compute ver by hashing S using the algorithm specified in in the 'hash' attribute (e.g., SHA-1 as defined in &rfc3174;). The hashed data MUST be generated with binary output and encoded using Base64 as specified in Section 4 of &rfc4648; (note: the Base64 output MUST NOT include whitespace and MUST set padding bits to zero). The OpenSSL command for producing such output with SHA-1 is is "echo -n 'S' | openssl dgst -binary -sha1 | openssl enc -nopad -base64".

For example, consider an entity whose service discovery category is "client", whose service discovery type is "pc", and whose supported features are "http://jabber.org/protocol/disco#info", "http://jabber.org/protocol/disco#items", and "http://jabber.org/protocol/muc". Using the SHA-1 algorightm, the value of the 'ver' attribute would be generated as follows:

1. S = ''
2. Only one identity: "client/pc"
3. S = 'client/pc<'
4. Sort the features: "http://jabber.org/protocol/disco#info", "http://jabber.org/protocol/disco#items", "http://jabber.org/protocol/muc".
5. S = 'client/pc<http://jabber.org/protocol/disco#info<http://jabber.org/protocol/disco#items<http://jabber.org/protocol/muc<'
6. ver = 8RovUdtOmiAjzj+xI7SK5BCw3A8=

A server that is managing an connected client's presence session MAY optimize presence notification traffic sent through the server by stripping off redundant capabilities annotations. Because of this, receivers of presence notifications MUST NOT expect an annotation on every presence notification they receive. If the server performs caps optimization, it MUST ensure that the first presence notification each subscriber receives contains the annotation. The server MUST also ensure that any changes in the caps infomration (e.g., an updated 'ver' attribute) are sent to all subscribers.

If a connected client determines that its server supports caps optimization, MAY choose to send the capabilities annotation only on the first presence packet, as well as whenever its capabilities change.

If an entity supports the entity capabilities protocol, it MUST advertise that fact by returning a feature of 'http://jabber.org/protocol/caps' in response to a service discovery information request.

If a server supports the Server Optimization functionality, it MUST also return a feature of 'http://jabber.org/protocol/caps#optimize' in response to service discovery information requests.

If two entities exchange messages but they do not normally exchange presence (i.e., via presence subscription), the entities MAY choose to send directed presence to each other, where the presence information SHOULD be annotated with the same capabilities information as each entity sends in broadcasted presence. Until and unless capabilities information has not been received from another entity, an application MUST assume that the other entity does not support capabilities.

It is RECOMMENDED for an application that processes entity capabilities information to cache associations between the 'ver' attribute and discovered features within the scope of one presence session. This obviates the need for extensive service discovery requests within a session.

It is OPTIONAL for an application to cache associates across presence sessions. However, since this obviates the need for extensive service discovery requests at the beginning of a session, such caching is strongly encouraged, especially in bandwidth-constrained environments.

Use of the protocol specified in this document might make some client-specific forms of attack slightly easier, since the attacker could more easily determine the type of client being used. However, since most clients respond to Service Discovery and Software Version requests without performing access control checks, there is no new vulnerability. Entities that wish to restrict access to capabilities information SHOULD use &xep0016; to define appropriate communications blocking (e.g., an entity MAY choose to allow IQ requests only from "trusted" entities, such as those with whom it has a presence subscription of "both"); note, however, that such restrictions may be incompatible with the recommendation regarding Directed Presence.

Adherence to the algorithm defined in the Generation of the ver Attribute section of this document for both generation and checking of the 'ver' attribute helps to guard against poisoning of entity capabilities information by malicious or improperly implemented entities.

If the value of the 'ver' attribute is a hash as defined herein (i.e., if the 'ver' attribute is not generated according to the Legacy Format), inclusion of the 'hash' attribute is required. Knowing explicitly that the value of the 'ver' attribute is a hash enables the recipient to avoid spurious notification of invalid hashes.

Theoretically it may become possible to launch a "pre-image" attack (see &rfc4270;) against the hashes used in the 'ver' attribute, at least when the SHA-1 algorithm is used. However, such attacks are not currently practical, and may not become practical in the foreseeable future. If and when such attacks become practical, this specification will be updated to strongly recommend use of a hashing algorithm that is safer than SHA-1, such as SHA-256. Nevertheless, the SHA-256 algorithm can be used today if implementors are concerned about the safety of the SHA-1 algorithm.

The 'name' attribute of the service discovery <identity/> element is not included in the hash generation method. The primary reason for excluding it is that it is human-readable text and therefore may be provided in different localized versions. As a result, its inclusion would needlessly multiply the number of possible hash values and thus the time and resources required to validate values of the 'ver' attribute.

This document requires no interaction with &IANA;.

Before Version 1.4 of this specification, the 'ver' attribute was generated differently, the 'ext' attribute was used more extensively, and the 'hash' and 'v' attributes were absent. For historical purposes, Version 1.3 of this specification is archived at <http://www.xmpp.org/extensions/attic/xep-0115-1.3.html>. For backwards-compatibility with the legacy format, the 'node' attribute is REQUIRED and the 'ext' attribute MAY be included.

An application can determine if the legacy format is in use by checking for the presence of the 'hash' attribute, which is REQUIRED in the current format.

If a caps-processing application supports the legacy format, it SHOULD check the 'node', 'ver', and 'ext' combinations as specified in the archived version 1.3 of this specification, and MAY cache the results.

If a caps-processing application does not support the legacy format, it SHOULD ignore the 'ver' value entirely (since the value cannot be verified) and SHOULD NOT cache it, since the application cannot validate the identity and features by checking the hash.

Thanks to Rachel Blackman, Dave Cridland, Richard Dobson, Olivier Goffart, Sergei Golovan, Justin Karneges, Ian Paterson, Kevin Smith, Tomasz Sterna, Michal Vaner, and Matt Yacobucci for comments and suggestions.