IRCv3 Message Tags

Copyright © 2012-2014 Alexey Sokolov <alexey-irc@asokolov.org>

Copyright © 2012 Stéphan Kochen <stephan@kochen.nl>

Copyright © 2012 Kyle Fuller <inbox@kylefuller.co.uk>

Copyright © 2016 Kiyoshi Aman <kiyoshi.aman@gmail.com>

Copyright © 2016-2019 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.


Introduction

Message tags are a mechanism for adding additional metadata on a per-message basis. This is achieved via an extension to the protocol message format, enabled via capability negotiation.

Tags are simple keys that MAY have optional string data as values.

Tagged messages can be sent by both servers and clients. The usage of individual tags are specified in their own documents.

Format

The message pseudo-BNF, as defined in RFC 1459, section 2.3.1 is extended as follows:

<message>       ::= ['@' <tags> <SPACE>] [':' <prefix> <SPACE> ] <command> <params> <crlf>
<tags>          ::= <tag> [';' <tag>]*
<tag>           ::= <key> ['=' <escaped_value>]
<key>           ::= [ <client_prefix> ] [ <vendor> '/' ] <key_name>
<client_prefix> ::= '+'
<key_name>      ::= <non-empty sequence of ascii letters, digits, hyphens ('-')>
<escaped_value> ::= <sequence of zero or more utf8 characters except NUL, CR, LF, semicolon (`;`) and SPACE>
<vendor>        ::= <host>

The ordering of tags is not meaningful.

Individual tag keys MUST only be used a maximum of once per message. Implementations receiving messages with more than one occurrence of a tag key name SHOULD disregard all but the final occurrence.

Implementations MUST treat tag key names as opaque identifiers and MUST NOT perform any validation that would reject the message if an invalid tag key name is used. This allows future modifications to the tag key name format.

Tag values MUST be encoded as UTF8. This ensures a shared interoperable baseline for data exchange. If tag values are encountered that cannot be decoded as UTF8, implementations MAY drop the value entirely but SHOULD NOT substitute replacement bytes in place of invalid data, which can result in collisions.

Implementations MUST interpret empty tag values (e.g. foo=) as equivalent to missing tag values (e.g. foo). Specifications MUST NOT differentiate meaning between tags with empty and missing values. Implementations MAY normalise tag values by converting the empty form to the missing form, but MUST NOT convert values from missing to empty, to prevent size limit issues.

Escaping values

The mapping between characters in tag values and their representation in <escaped value> is defined as follows:

Character Sequence in <escaped value>
; (semicolon) \: (backslash and colon)
SPACE \s
\ \\
CR \r
LF \n
all others the character itself

This escape format is space efficient, and ensures that message parts can easily be split on spaces and semi-colons before further parsing.

If a lone \ exists at the end of an escaped value (with no escape character following it), then there SHOULD be no output character. For example, the escaped value test\ should unescape to test. If a \ exists with no valid escape character (for example, \b), then the invalid backslash SHOULD be dropped. For example, \b should unescape to just b.

Capabilities

Tags are enabled via capability negotiation. Clients and servers that negotiate a capability that uses tags MUST support the full tag format above.

The following capabilities implicitly depend on the tag format:

But there is also a generic capability for negotiating tag usage.

message-tags capability

The message-tags capability allows clients to indicate support for receiving any well-formed tag, whether or not it is recognised or used. It also frees servers from having to filter individual message tags for each client response.

The message-tags capability is the only guarantee that an implementation supports the full tag format. Following the robustness principle, implementations that use tags MUST support the full message tag format. However, implementations SHOULD NOT assume support for all tags when an individual tag capability is negotiated.

This capability enables the use of client-only tags and the TAGMSG command, described below.

Clients requesting this capability indicate that they are capable of parsing all well-formed tags, even if they don’t handle them specifically.

Servers advertising this capability indicate that they are capable of parsing any tag received from a client, with or without the client-only prefix.

Client-only tags

Client-only tags are tags normally sent by clients that servers MUST attach as-is to any relevant message relayed to other clients. A client-only tag is prefixed with a plus sign (+). Servers MAY send client-only tags that weren’t provided explicitly by the client.

The client-only tag prefix allows servers to safely relay untrusted client tags, keeping them distinct from unprefixed tags that carry verified meaning.

Client-only tags MUST be relayed on PRIVMSG, NOTICE and TAGMSG messages, and MAY be relayed on other messages.

Any server-sent tags attached to messages MUST be included before client-only tags to prevent them from being pushed outside of the byte limit.

The expected client behaviour of individual client-only tags are defined in separate specifications.

This means client-only tags that aren’t specified in the IRCv3 extension registry MUST use a vendor prefix and SHOULD be submitted to the IRCv3 working group for consideration if they are appropriate for more widespread adoption. See Rules for naming message tags.

Client-only tags are intended to replace the use of future CTCP commands.

The TAGMSG tag-only message

   Command: TAGMSG
Parameters: <msgtarget>

A new message command TAGMSG is defined for sending messages with tags but no text content. This message MUST be delivered to targets in the same way as PRIVMSG and NOTICE messages. This means for example, honouring channel membership, modes, echo-message, STATUSMSG prefixes, etc.

Servers MAY apply moderation to this command using existing or newly specified modes or configuration.

Servers MUST NOT deliver TAGMSG to clients that haven’t negotiated the message tags capability.

See PRIVMSG in RFC2812 for more details on replies and examples.

Clients that receive a TAGMSG command MUST NOT display them in the message history by default. Display guidelines are defined in the specifications of tags attached to the message.

Rules for naming message tags

Full tag names, including any vendor prefixes MUST be treated as an opaque identifier.

There are two tag namespaces:

Vendor-Specific

Names which contain a slash character (/) designate a vendor-specific tag 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 tags should be submitted to the IRCv3 working group for consideration.

Drafts

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

While tags are in draft status, they may need to be given a new identifier, to prevent implementation compatibility issues. When updating a draft tag key 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.

Standardized

Reserved names for which a corresponding document exists in the IRCv3 Extension Registry.

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 tag then you MUST use a vendor-specific name (see above).

Size limit

The size limit for message tags is 8191 bytes, including the leading '@' (0x40) and trailing space ' ' (0x20) characters. The size limit for the rest of the message remains unchanged at 512 bytes.

This limit is separated between tags added by the server and tags sent by the client. This prevents servers from overflowing the overall limit by adding tags to a client message sent within the allowed limit.

In the following description, tag data describes the bytes between the leading and trailing tag separators mentioned above.

Clients MUST NOT send messages with tag data exceeding 4094 bytes, this includes tags with or without the client-only prefix.

Servers MUST NOT add tag data exceeding 4094 bytes to messages.

<server_max>   (4096)  :: '@' <tag_data 4094> ' '
<client_max>   (4096)  :: '@' <tag_data 4094> ' '
<combined_max> (8191)  :: '@' <tag_data 4094> ';' <tag_data 4094> ' '

Servers MUST reply with the ERR_INPUTTOOLONG (417) error numeric if a client sends a message with more tag data than the allowed limit. Servers MUST NOT truncate tags and MUST always reject lines that exceed this limit.

417    ERR_INPUTTOOLONG
      ":Input line was too long"

If a server sends a message with more tag data than the allowed limit, clients MAY ignore the message.

Security considerations

Client-only tags should be treated as untrusted data. They can contain any value and are not validated by servers in any way. The server MAY unescape tag values after receiving data and escape those values before sending them out.

Tags without the client-only prefix MUST be removed by the server before being relayed with any message to another client.

Some specifications may involve servers accepting client-sent tags without the client-only prefix. These specifications MUST define a process to be performed by the server on these tags prior to their removal.

Client implementation considerations

There is no guarantee that other clients will support or display the client-only tags they receive, so these tags should only be used to enhance messages with non-critical information. Messages should still make sense to clients with no support for tags.

Moderation considerations

This section is non-normative.

Moderation tools available to channel and network operators typically operate on message text and sender information only. To avoid abusive content being sent in client-only tags, servers, services and management bot implementations may wish to enable moderation features that act on tag content as well.

Examples

This section is non-normative. The tags used in these examples may or may not have a specified meaning elsewhere.


A message with 0 tags:

:nick!ident@host.com PRIVMSG me :Hello

A message sent by the server with 3 tags:

S: @aaa=bbb;ccc;example.com/ddd=eee :nick!ident@host.com PRIVMSG me :Hello

A message sent by a client with the example-tag tag:

C: @example-tag=example-value PRIVMSG #channel :Message

A message sent by a client with the +example-client-tag client-only tag:

C: @+example-client-tag=example-value PRIVMSG #channel :Message

In this example

The server sends these messages:

S: :nick!user@example.com PRIVMSG #channel :https://example.com/a-news-story
S: @+icon=https://example.com/favicon.png :url_bot!bot@example.com PRIVMSG #channel :Example.com: A News Story

An example of a vendor-prefixed client-only tag:

C: @+example.com/foo=bar :irc.example.com NOTICE #channel :A vendor-prefixed client-only tagged message

A client-only tag +example with a value containing valid raw and escaped characters:

In this example, plus signs, colons, equals signs and commas are transmitted raw in tag values; while semicolons, spaces and backslashes are escaped.

C: @+example=raw+:=,escaped\:\s\\ :irc.example.com NOTICE #channel :Message

A TAGMSG sent by a client with the +example-client-tag client-only tag:

C: @+example-client-tag=example-value TAGMSG #channel

A TAGMSG sent by a client to channel ops with a client-only tag and STATUSMSG prefix:

C: @+example-client-tag=example-value TAGMSG @#channel

A TAGMSG sent to a channel with labeled-response and client-only tags. The message is returned by the server with a msgid tag via labeled echo-message to the originating client and as normal to other clients in the channel.

C1-S: @label=123;+example-client-tag=example-value TAGMSG #channel
S-C1: @label=123;msgid=abc;+example-client-tag=example-value :nick!user@example.com TAGMSG #channel
S-C2: @msgid=abc;+example-client-tag=example-value :nick!user@example.com TAGMSG #channel

A TAGMSG sent by a client with tags that exceed the size limit. The message is rejected by the server with an ERR_INPUTTOOLONG (417) error numeric. [...] is used to represent tags omitted for readability.

C: @+tag1;+tag2;+tag[...];+tag5000 TAGMSG #channel
S: :server.example.com 417 nick :Input line was too long

A TAGMSG sent by a client with an un-prefixed tag that has no specified behaviour. The server removes the tag before relaying the message

C: @unknown-tag TAGMSG #channel
S: :nick!user@example.com TAGMSG #channel

Errata

Previous versions of this spec did not specify that the full tag name MUST be parsed as an opaque identifier. This was added to improve client resiliency.

Previous versions of this spec did not specify how to handle trailing backslashes with no escape character. This was added to help consistency across implementations.

Previous versions of this spec did not specify how to handle invalid escapes. This was clarified to help consistency across implementations.

Previous versions of this spec did not specify the difference between empty and missing tag values.

Previous versions of this spec did not specify the UTF8 encoding for tag values


Software supporting message-tags: IRCCloud Teams, InspIRCd, Oragono, AdiIRC, Irssi, IRCCloud, The Lounge, IRCCloud (as Server), Moon Moon, BitBot, girc, Kitteh IRC Client Library