Client Capability Negotiation

Copyright © 2004-2012 Kevin L. Mitchell <klmitch@mit.edu>

Copyright © 2004-2012 Perry Lorier <isomer@undernet.org>

Copyright © 2004-2012 Lee Hardy <lee@leeh.co.uk>

Copyright © 2009-2012 William Pitcock <nenolod@dereferenced.org>

Copyright © 2014 Attila Molnar <attilamolnar@hush.com>

Copyright © 2017-2019 Daniel Oakley <daniel@danieloaks.net>

Copyright © 2017-2019 James Wheare <james@irccloud.com>

Unlimited redistribution and modification of this document is allowed provided that the above copyright notice and this permission notice remains intact.


What Client Capability Negotiation attempts to solve 🔗

Client Capability Negotiation allows IRC clients and servers to negotiate new features in a backwards-compatible way – even features that change how the protocol works in deep and extensive ways. Capability negotiation avoids the issues of breaking compatibility with clients/servers not supporting new features by allowing clients and servers to enable the features they both understand.

Capability Negotiation means that client and server authors can develop new extensions to the protocol, and software can use (or not use them) as they wish. It allows older clients to connect to servers supporting new features and vice-versa.

Connection Registration 🔗

Upon connecting to the IRC server, clients need a way to negotiate capabilities (protocol extensions) with the server. Negotiation is done with the CAP command, using the subcommands below to list and request capabilities.

Upon connecting to the server, the client attempts to start the capability negotiation process, as well as sending the standard NICK/USER commands (to complete registration if the server doesn’t support capability negotiation).

Upon connecting to the IRC server, clients SHOULD send one of the following messages:

Following this, the client MUST send the standard NICK and USER IRC commands.

Upon receiving either a CAP LS or CAP REQ command during connection registration, the server MUST not complete registration until the client sends a CAP END command to indicate that capability negotiation has ended. This allows clients to request their desired capabilities before completing registration.

Once capability negotiation has completed with a client-sent CAP END command, registration continues as normal.

With the above registration process, clients will either receive a CAP message indicating that registration is paused while capability negotiation happens, or registration will complete immediately, indicating that the server doesn’t support capability negotiation.

Example with capability negotiation:

Client: CAP LS 302
Client: NICK dan
Client: USER d * 0 :This is a really good name
Server: CAP * LS :multi-prefix sasl
Client: CAP REQ :multi-prefix
Server: CAP * ACK multi-prefix
Client: CAP END
Server: 001 dan :Welcome to the Internet Relay Network dan
...

Example with capability negotiation, but where the client recognises no advertised caps:

Client: CAP LS 302
Client: NICK dan
Client: USER d * 0 :This is a really good name
Server: CAP * LS :draft/example-1 draft/example-2
Client: CAP END
Server: 001 dan :Welcome to the Internet Relay Network dan
...

Example of capability negotiation without a prior LS:

Client: CAP REQ multi-prefix
Client: NICK dan
Client: USER d * 0 :This is a really good name
Server: CAP * ACK multi-prefix
Client: CAP END
Server: 001 dan :Welcome to the Internet Relay Network dan

Example where the server doesn’t support capability negotiation:

Client: CAP LS 302
Client: NICK dan
Client: USER d * 0 :This is a really good name
Server: 001 dan :Welcome to the Internet Relay Network dan
...

Example where the client doesn’t support capability negotiation:

Client: NICK dan
Client: USER d * 0 :This is a really good name
Server: 001 dan :Welcome to the Internet Relay Network dan
...

The CAP command 🔗

The client capability negotiation extension is implemented by the addition of one command with several subcommands. The command added is named CAP. CAP takes a single required subcommand, optionally followed by a single parameter. Each subcommand defines any further parameters.

The format of a client-sent CAP message is:

CAP <subcommand> [<parameters>...]

The format of a server-sent CAP message is:

CAP <nick> <subcommand> [<parameters>...]

The subcommands for CAP are: LS, LIST, REQ, ACK, NAK, NEW, DEL, and END.

Here is a handy table of the available CAP subcommands and which side sends them:

Subcommand Client sent Server sent (triggered by)
LS * * (LS)
LIST * * (LIST)
REQ *  
ACK   * (REQ)
NAK   * (REQ)
END *  
NEW (302)   *
DEL (302)   *

Some subcommands may include a space-separated list of capabilities as their final parameter. The list of capabilities MUST be parsed and processed from left to right and capabilities SHOULD only be sent once per command. If a capability with values is sent multiple times, the last one received takes priority. Clients MUST ignore any trailing whitespace.

If a client sends a subcommand which is not in the list above or otherwise issues an invalid command, then numeric 410 (ERR_INVALIDCAPCMD) should be sent. The first parameter after the client identifier (usually nickname) should be the commandname; the second parameter should be a human-readable description of the error.

Replies from the server must contain the client identifier name or asterisk if one is not yet available.

Example with nick jw:

Client: CAP FOO
Server: :example.org 410 jw FOO :Invalid CAP command

Example before nick is set (e.g. during registration):

Client: CAP FOO
Server: :example.org 410 * FOO :Invalid CAP command

The server MUST accept the CAP command at any time, including after registration.

The CAP LS subcommand 🔗

The LS subcommand is used to list the capabilities supported by the server. The client should send an LS subcommand with no other arguments to solicit a list of all capabilities.

The format of a server-sent CAP LS message is:

CAP LS [<version>]

The format of a server-sent CAP LS message is:

CAP <nick> LS :[<extension 1> ... [<extension n>]]

The format of a server-sent CAP LS message, with CAP LS 302 is:

CAP <nick> LS [*] :[<extension 1> ... [<extension n>]]

If a server receives an LS subcommand while client registration is in progress, it MUST suspend registration until an END subcommand is received from the client.

When sent by the server, the last parameter is a space-separated list of capabilities (possibly with values, depending on the CAP LS Version described below). If no capabilities are available, an empty parameter MUST be sent.

Example:

Client: CAP LS
Server: CAP * LS :multi-prefix sasl

Example with no available capabilities:

Client: CAP LS
Server: CAP * LS :

CAP LS Version 🔗

The LS subcommand has an additional argument which is the version number of the latest capability negotiation protocol supported by the client. Newer versions of capability negotiation allow newer features, as described below:

Clients that send 302 as the CAP LS version are presumed to support CAP LS 302 features for the future life of the connection. Clients that do not send any version number with CAP LS are presumed to not support these extra features.

If a client has not indicated support for CAP LS 302 features, the server MUST NOT send these new features to the client.

When CAP version 302 is enabled, the client also implicitly indicates support for the cap-notify capability listed below, and support for the relevant NEW and DEL subcommands.

The CAP version number argument MUST be treated as a number by the server. If the client’s CAP version is not supported by the server, the server MUST enable the features it does support below the given version. For example, if a server supports 302 and 303 features, and a client sends version 304, the server would enable those 302 and 303 features (just as it would do if a client sent the 303 version number). However, if the server also supported a 306 feature, it would NOT enable this feature for the client.

If a client sends a higher CAP version at any time, the server MUST store the higher version. If a client sends a lower CAP version (or omits the version number entirely), servers SHOULD return a CAP LS reply consistent with the request’s version, but keep storing the original (higher) version.

Example:

Client: CAP LS 302
Server: CAP * LS :multi-prefix sasl=PLAIN,EXTERNAL
Client: CAP LS
Server: CAP * LS :multi-prefix sasl
... server should continue considering the client to support CAP LS version 302 ...

Example with a server supporting only 302 versioned features:

Client: CAP LS 307
Server: CAP * LS :multi-prefix sasl=PLAIN,EXTERNAL
CAP LS Version Features 🔗

As an overview, these are the new features introduced with each CAP LS version:

CAP Name Description
302 Capability values Additional data with each capability name when advertised in CAP LS and CAP NEW.
302 Multiline replies CAP LS and CAP LIST can be split across multiple lines, with a minor syntax change that allows clients to wait for the last message and process them together.
302 cap-notify This capability is enabled implicitly with 302, and adds the CAP NEW and CAP DEL messages which let the client know about added and removed capabilities.

Capability Values 🔗

If the client supports CAP version 302, the server MAY specify additional data for each capability using the <name>=<value> format in CAP LS and CAP NEW replies, but MUST NOT specify additional data for any other CAP subcommands.

Each capability, if it supports a value, defines what this value means in its specification.

Example:

Client: CAP LS 302
Server: CAP * LS :multi-prefix sasl=PLAIN,EXTERNAL server-time draft/packing=EX1,EX2

Multiline replies to CAP LS and CAP LIST 🔗

If the client supports CAP version 302, the server MAY send multiple lines in response to CAP LS and CAP LIST. Clients that support CAP version 302 MUST handle the continuation format described as follows:

If the reply contains multiple lines (due to IRC line length limitations), and the client supports CAP version 302, all but the last reply MUST have a parameter containing only an asterisk (*) preceding the capability list. This lets clients know that more CAP lines are incoming, so that it can delay capability negotiation until it has seen all available server capabilities.

Example of a multiline LS reply to a client that supports CAP version 302:

Client: CAP LS 302
Server: CAP * LS * :multi-prefix extended-join account-notify batch invite-notify tls
Server: CAP * LS * :cap-notify server-time example.org/dummy-cap=dummyvalue example.org/second-dummy-cap
Server: CAP * LS :userhost-in-names sasl=EXTERNAL,DH-AES,DH-BLOWFISH,ECDSA-NIST256P-CHALLENGE,PLAIN

Example of a multiline LIST reply to a client that supports CAP version 302:

Client: CAP LIST
Server: CAP modernclient LIST * :example.org/example-cap example.org/second-example-cap account-notify
Server: CAP modernclient LIST :invite-notify batch example.org/third-example-cap

The CAP LIST subcommand 🔗

The LIST subcommand is used to list the capabilities enabled on the client’s connection. The client should send a LIST subcommand with no other arguments to solicit a list of enabled capabilities.

The format of a client-sent CAP LIST message is:

CAP LIST

The format of a server-sent CAP LIST message is:

CAP <nick> LIST :[<extension 1> ... [<extension n>]]

The format of a server-sent CAP LIST message, with CAP LIST 302 is:

CAP <nick> LIST [*] :[<extension 1> ... [<extension n>]]

When sent by the server, the last parameter is a space-separated list of capabilities. If no capabilities are enabled, an empty parameter must be sent.

Example:

Client: CAP LIST
Server: CAP * LIST :multi-prefix

Example with no enabled capabilities:

Client: CAP LIST
Server: CAP * LIST :

The CAP REQ subcommand 🔗

The REQ subcommand is used to request a change in capabilities associated with the active connection. The last parameter is a space-separated list of capabilities. Each capability identifier may be prefixed with a dash (-) to designate that the capability should be disabled.

The format of a CAP REQ message is:

CAP REQ :<extension 1> [[-]<extension 2> ... [[-]<extension n>]]

If a client requests a capability which is already enabled, or tries to disable a capability which is not enabled, the server MUST continue processing the REQ subcommand as though handling this capability was successful.

The capability identifier set must be accepted as a whole, or rejected entirely.

Clients SHOULD ensure that their list of requested capabilities is not too long to be replied to with a single ACK or NAK message. If a REQ’s final parameter gets sufficiently large (approaching the 510 byte limit), clients SHOULD instead send multiple REQ subcommands.

If a server receives a REQ subcommand while client registration is in progress, it MUST suspend registration until an END subcommand is received.

Example adding a capability:

Client: CAP REQ :multi-prefix sasl
Server: CAP * ACK :multi-prefix sasl

Example removing a capability:

Client: CAP REQ :-userhost-in-names
Server: CAP * ACK :-userhost-in-names

The CAP ACK subcommand 🔗

The ACK subcommand is sent by the server to acknowledge a client-sent REQ, and let the client know that their requested capabilities have been enabled.

The format of a CAP ACK message is:

CAP <nick> ACK :[-]<extension 1> [[-]<extension 2> ... [[-]<extension n>]]]

The last parameter is a space-separated list of capabilities. Each capability name may be prefixed with a dash (-), indicating that this capability has been disabled as requested.

If an ACK reply originating from the server is spread across multiple lines, a client MUST NOT change capabilities until the last ACK of the set is received. Equally, a server MUST NOT change the capabilities of the client until the last ACK of the set has been sent.

The CAP NAK subcommand 🔗

The NAK subcommand designates that the requested capability change was rejected. The server MUST NOT make any change to any capabilities if it replies with a NAK subcommand.

The format of a CAP NAK message is:

CAP <nick> NAK :<extension 1> [<extension 2> ... [<extension n>]]

The last parameter is a space-separated list of capabilities.

Example:

Client: CAP REQ :multi-prefix sasl ex3
Server: CAP * NAK :multi-prefix sasl ex3

The CAP END subcommand 🔗

The END subcommand signals to the server that capability negotiation is complete and requests that the server continue with client registration. If the client is already registered, this command MUST be ignored by the server.

The format of a CAP END message is:

CAP END

The CAP NEW subcommand 🔗

The NEW subcommand MUST ONLY be sent to clients that have negotiated CAP version 302 or enabled the cap-notify capability.

The NEW subcommand signals that the server supports one or more new capabilities, and may be sent at any time. Clients that support CAP NEW messages SHOULD respond with a CAP REQ message if they wish to enable one or more of the newly-offered capabilities.

The format of a CAP NEW message is:

CAP <nick> NEW :<extension 1> [<extension 2> ... [<extension n>]]

As with LS, the last parameter is a space-separated list of new capabilities that are now offered by the server. If the client supports CAP version 302, the capabilities SHOULD be listed with values, as in the CAP LS response.

Example:

Server: :irc.example.com CAP modernclient NEW :batch

Example with following REQ:

Server: :irc.example.com CAP tester NEW :away-notify extended-join
Client: CAP REQ :extended-join away-notify
Server: :irc.example.com CAP tester ACK :extended-join away-notify

The CAP DEL subcommand 🔗

The DEL subcommand MUST ONLY be sent to clients that have negotiated CAP version 302 or enabled the cap-notify capability.

The DEL subcommand signals that the server no longer supports one or more capabilities that have been advertised. Upon receiving a CAP DEL message, the client MUST treat the listed capabilities as cancelled and no longer available. Clients SHOULD NOT send CAP REQ messages to cancel the capabilities in CAP DEL, as they have already been cancelled by the server.

Servers MUST cancel any capability-specific behavior for a client after sending the CAP DEL message to the client.

Clients MUST gracefully handle situations when the server removes support for any capability.

The format of a CAP DEL message is:

CAP <nick> DEL :<extension 1> [<extension 2> ... [<extension n>]]

The last parameter is a space-separated list of capabilities that are no longer available.

Example:

Server: :irc.example.com CAP modernclient DEL :userhost-in-names multi-prefix away-notify

cap-notify 🔗

The cap-notify capability indicates support for the NEW and DEL messages listed above. This capability MUST be implicitly enabled if the client requests CAP LS with a version of 302 or newer (CAP LS 302), as described in the CAP LS section above.

Further, servers MUST NOT disable the cap-notify capability if the client requests CAP LS with a version of 302 or newer.

When implicitly enabled via this mechanism, servers MAY list the cap-notify capability in CAP LS and CAP LIST responses. Clients MAY also request the capability with CAP REQ, and capable servers MUST accept and CAP ACK the request without side effects. This lets clients blindly enable this capability, regardless of it being implicitly enabled by the server.

Clients that do not support CAP version 302 MAY request the cap-notify capability explicitly. Such clients MAY explicitly disable the capability, and servers MUST allow these clients to do so. This to ease the adaptation of new features.

When enabled, the server MUST notify the client about all new capabilities and about existing capabilities that are no longer available using the NEW and DEL subcommands listed above.

Rules for naming capabilities 🔗

The full capability name MUST be treated as an opaque identifier.

Capability names are case-sensitive, and MUST NOT start with a hyphen (-). Typical capability names SHOULD be lowercase, and use hyphens to separate words. For example: echo-message, extended-join, invite-notify, draft/labeled-response, message-tags.

There are different types of capability names, which are described below.

Vendor-Specific capabilities 🔗

Names which contain a slash character (/) designate a vendor-specific capability namespace.

These names are prefixed by a valid DNS domain name.

For example: znc.in/server-time.

In cases where the prefix contains non-ASCII characters, punycode MUST be used, e.g. xn--e1afmkfd.org/foo.

Vendor-Specific capabilities should be submitted to the IRCv3 working group for consideration.

Draft capabilities 🔗

The draft/ vendor namespace may be used when the working group is considering specifications. However, vendor names should be preferred.

While capabilities are in draft status, they may need to be given a new identifier, to prevent clients from negotiating incompatible versions. When updating a draft capability name, the typical method is to add -0.x to the name, where x is a version number. For example: draft/foo would become draft/foo-0.2, and so on.

Standardised capabilities 🔗

Standardised capabilities have no vendor namespace, and are listed on the IRCv3 Capability Registry. These capabilities also have one or more documents in the set of IRCv3 specifications describing how they work.

The IRCv3 Working Group reserves the right to reuse names which have not been submitted to the registry. If you do not wish to submit your capability then you MUST use a vendor-specific name (see above).

Errata 🔗

Previous versions of this specification referred to a CAP CLEAR command, which has been removed because it is not useful. We do not recommend implementing or supporting CAP CLEAR. See issue #134 for more information, including rationale for this clarification.

Previous versions of this spec did not include specific references to re-enabling or re-disabling a capability in the CAP REQ subcommand section. This was clarified in later versions of the specification.

Previous versions of this spec did not specify how to handle CAP LS when a server did not support any capabilities. This was clarified to match CAP LIST, requiring a reply with an empty parameter.

Previous versions of this spec did not specify that the full capability name MUST be treated as an opaque identifier. This was added to better suit real-world usage and to improve client resiliency.

Previous versions of this spec defined ‘capability modifiers’. These are not in use by software and have been removed from this specification.

Previous versions of this spec did not specify when the <name>=<value> format could be used. This was clarified to limit capability values to the LS and NEW subcommands.

Previous versions of this spec did not link to the cap-notify specification (nor note the fact that it is automatically enabled for clients that enable CAP version 302). The specification now includes the cap-notify spec and lets authors know about the implicit enabling of this capability.

Previous versions of this spec mistakenly missed <nick> between CAP and NEW/DEL subcommands, but had it in the examples anyway.

Previous versions of this spec did not mention whether NEW and DEL can have values or not.

Previous versions of this spec were missing clarification of client and server behaviour when the capability is implicitly enabled with CAP version 302 or newer.

Previous versions of this spec listed that CAP ACK could be sent from the client to server, for capabilities that required extra client acknowledgement. This was removed since cap modifiers have been deprecated and removed (except for -, which has been specified in other ways).

Previous versions of this spec listed connection registration in the middle. This section has been rewritten and placed near the front, now including examples of negotiating capabilities during initial registration.

Previous versions of this spec included capability modifiers. These have been deprecated and removed from this specification (and have been as of ~2015 with capability negotiation 3.2).

Previous versions of this spec recommended sending CAP END upon connection if the client didn’t want to perform CAP negotiation. This advice has been removed as sending only CAP END doesn’t make much sense in this case.

Previous versions of this spec required that when sending CAP NAK, servers MUST have the final parameter (the capability list) consist of “at least the first 100 characters of the capability list in the REQ subcommand which triggered the NAK”. This was removed as it’s a bit arbitrary, already covered in the existing language, and honestly more likely to just confuse vendors.

Previous versions of this spec did not mention how servers treat higher CAP LS versions. This was clarified so that if a client sends a higher version number than the server supports, they get the appropriate features.

Previous versions of this spec did not mention how servers handle clients attempting to downgrade their CAP LS version. It has been clarified that clients MAY NOT downgrade this.

Clarified that multiline LS and LIST replies must only be used for CAP 302.

Previous versions of this spec did not state that capability names MUST NOT start with a hyphen.

Previous versions of this spec did not state that space-separated capability lists may end with a trailing space.

Previous versions of this spec did not give formats for commands other than CAP NEW and CAP DEL.


Software supporting CAP: Ergo, IRCCloud Teams, ircd-hybrid, InspIRCd, Nefarious IRCu, Planio Team Chat, Solanum, txircd, UnrealIRCd, AdiIRC, Ambassador, BitchX, catgirl, Colloquy, glirc, Halloy, HexChat, IceChat, Irssi, Konversation, KVIrc, LimeChat, mIRC, Mozilla Thunderbird, Pidgin, Quassel, senpai, Srain, Swirc, Textual, WeeChat, gamja, IRCCloud, Kiwi IRC, The Lounge, Mibbit, PIRC.pl web client, IRC for Android, Colloquy, CoreIRC, LimeChat, Palaver, Quasseldroid, Goguma, IRCCloud (as Server), KiwiBNC (as Server), KiwiBNC (as Client), pounce (as Server), pounce (as Client), soju (as Server), soju (as Client), ZNC (as Server), ZNC (as Client), BitBot, Eggdrop, Limnoria, Moon Moon, Sopel (ex Willie), ChatSharp, Communi, girc, irc-framework, ircrobots, Kitteh IRC Client Library, Net::Async::IRC, pydle, Rust irc, Warren, zIRC, BitlBee, Heisenbridge (bouncer mode), matrix-appservice-irc, Matrix2051, Matterbridge, PyLink (clientbot mode)

Software supporting CAP 302: Ergo, IRCCloud Teams, ircd-hybrid, InspIRCd, Solanum, txircd, UnrealIRCd, AdiIRC, Ambassador, Colloquy, glirc, Halloy, HexChat, IceChat, Irssi, Konversation, mIRC, Mozilla Thunderbird, Quassel, senpai, Srain, Swirc, Textual, WeeChat, gamja, IRCCloud, Kiwi IRC, The Lounge, PIRC.pl web client, Colloquy, CoreIRC, Palaver, Quasseldroid, Goguma, IRCCloud (as Server), KiwiBNC (as Server), KiwiBNC (as Client), pounce (as Server), pounce (as Client), soju (as Server), soju (as Client), ZNC (as Server), ZNC (as Client), BitBot, Eggdrop, Limnoria, Moon Moon, Sopel (ex Willie), ChatSharp, Communi, girc, irc-framework, ircrobots, Kitteh IRC Client Library, pydle, Rust irc, Warren, zIRC, BitlBee, matrix-appservice-irc, Matrix2051, Matterbridge, PyLink (clientbot mode)

Software supporting cap-notify: Ergo, IRCCloud Teams, ircd-hybrid, InspIRCd, Solanum, txircd, UnrealIRCd, AdiIRC, Ambassador, Colloquy, glirc, Halloy, HexChat, IceChat, Irssi, Konversation, mIRC, Mozilla Thunderbird, Quassel, senpai, Srain, Swirc, WeeChat, gamja, IRCCloud, Kiwi IRC, The Lounge, PIRC.pl web client, Colloquy, CoreIRC, Palaver, Quasseldroid, Goguma, IRCCloud (as Server), KiwiBNC (as Server), KiwiBNC (as Client), pounce (as Server), pounce (as Client), soju (as Server), soju (as Client), ZNC (as Server), ZNC (as Client), BitBot, Eggdrop, Limnoria, Moon Moon, Sopel (ex Willie), ChatSharp, Communi, girc, irc-framework, ircrobots, Kitteh IRC Client Library, Rust irc, Warren, zIRC, PyLink (clientbot mode)