blob: d806b8eaf59bfdee0469eb24ab44be0c6dfa0ddf [file] [log] [blame]
<?xml version="1.0" standalone="no" ?>
<!DOCTYPE article PUBLIC "-//OASIS//DTD DocBook XML V4.1.2//EN"
"http://www.oasis-open.org/docbook/xml/4.1.2/docbookx.dtd"
[
]>
<article id="index">
<articleinfo>
<title>D-Bus Specification</title>
<releaseinfo>Version 0.19</releaseinfo>
<date>2012-02-21</date>
<authorgroup>
<author>
<firstname>Havoc</firstname>
<surname>Pennington</surname>
<affiliation>
<orgname>Red Hat, Inc.</orgname>
<address>
<email>hp@pobox.com</email>
</address>
</affiliation>
</author>
<author>
<firstname>Anders</firstname>
<surname>Carlsson</surname>
<affiliation>
<orgname>CodeFactory AB</orgname>
<address>
<email>andersca@codefactory.se</email>
</address>
</affiliation>
</author>
<author>
<firstname>Alexander</firstname>
<surname>Larsson</surname>
<affiliation>
<orgname>Red Hat, Inc.</orgname>
<address>
<email>alexl@redhat.com</email>
</address>
</affiliation>
</author>
<author>
<firstname>Sven</firstname>
<surname>Herzberg</surname>
<affiliation>
<orgname>Imendio AB</orgname>
<address>
<email>sven@imendio.com</email>
</address>
</affiliation>
</author>
<author>
<firstname>Simon</firstname>
<surname>McVittie</surname>
<affiliation>
<orgname>Collabora Ltd.</orgname>
<address>
<email>simon.mcvittie@collabora.co.uk</email>
</address>
</affiliation>
</author>
<author>
<firstname>David</firstname>
<surname>Zeuthen</surname>
<affiliation>
<orgname>Red Hat, Inc.</orgname>
<address>
<email>davidz@redhat.com</email>
</address>
</affiliation>
</author>
</authorgroup>
<revhistory>
<revision>
<revnumber>current</revnumber>
<date><ulink url='http://cgit.freedesktop.org/dbus/dbus/log/doc/dbus-specification.xml'>commit log</ulink></date>
<authorinitials></authorinitials>
<revremark></revremark>
</revision>
<revision>
<revnumber>0.19</revnumber>
<date>20 February 2012</date>
<authorinitials>smcv/lp</authorinitials>
<revremark>formally define unique connection names and well-known
bus names; document best practices for interface, bus, member and
error names, and object paths; document the search path for session
and system services on Unix; document the systemd transport</revremark>
</revision>
<revision>
<revnumber>0.18</revnumber>
<date>29 July 2011</date>
<authorinitials>smcv</authorinitials>
<revremark>define eavesdropping, unicast, broadcast; add eavesdrop
match keyword; promote type system to a top-level section</revremark>
</revision>
<revision>
<revnumber>0.17</revnumber>
<date>1 June 2011</date>
<authorinitials>smcv/davidz</authorinitials>
<revremark>define ObjectManager; reserve extra pseudo-type-codes used
by GVariant</revremark>
</revision>
<revision>
<revnumber>0.16</revnumber>
<date>11 April 2011</date>
<authorinitials></authorinitials>
<revremark>add path_namespace, arg0namespace; argNpath matches object
paths</revremark>
</revision>
<revision>
<revnumber>0.15</revnumber>
<date>3 November 2010</date>
<authorinitials></authorinitials>
<revremark></revremark>
</revision>
<revision>
<revnumber>0.14</revnumber>
<date>12 May 2010</date>
<authorinitials></authorinitials>
<revremark></revremark>
</revision>
<revision>
<revnumber>0.13</revnumber>
<date>23 Dezember 2009</date>
<authorinitials></authorinitials>
<revremark></revremark>
</revision>
<revision>
<revnumber>0.12</revnumber>
<date>7 November, 2006</date>
<authorinitials></authorinitials>
<revremark></revremark>
</revision>
<revision>
<revnumber>0.11</revnumber>
<date>6 February 2005</date>
<authorinitials></authorinitials>
<revremark></revremark>
</revision>
<revision>
<revnumber>0.10</revnumber>
<date>28 January 2005</date>
<authorinitials></authorinitials>
<revremark></revremark>
</revision>
<revision>
<revnumber>0.9</revnumber>
<date>7 Januar 2005</date>
<authorinitials></authorinitials>
<revremark></revremark>
</revision>
<revision>
<revnumber>0.8</revnumber>
<date>06 September 2003</date>
<authorinitials></authorinitials>
<revremark>First released document.</revremark>
</revision>
</revhistory>
</articleinfo>
<sect1 id="introduction">
<title>Introduction</title>
<para>
D-Bus is a system for low-latency, low-overhead, easy to use
interprocess communication (IPC). In more detail:
<itemizedlist>
<listitem>
<para>
D-Bus is <emphasis>low-latency</emphasis> because it is designed
to avoid round trips and allow asynchronous operation, much like
the X protocol.
</para>
</listitem>
<listitem>
<para>
D-Bus is <emphasis>low-overhead</emphasis> because it uses a
binary protocol, and does not have to convert to and from a text
format such as XML. Because D-Bus is intended for potentially
high-resolution same-machine IPC, not primarily for Internet IPC,
this is an interesting optimization.
</para>
</listitem>
<listitem>
<para>
D-Bus is <emphasis>easy to use</emphasis> because it works in terms
of <firstterm>messages</firstterm> rather than byte streams, and
automatically handles a lot of the hard IPC issues. Also, the D-Bus
library is designed to be wrapped in a way that lets developers use
their framework's existing object/type system, rather than learning
a new one specifically for IPC.
</para>
</listitem>
</itemizedlist>
</para>
<para>
The base D-Bus protocol is a one-to-one (peer-to-peer or client-server)
protocol, specified in <xref linkend="message-protocol"/>. That is, it is
a system for one application to talk to a single other
application. However, the primary intended application of the protocol is the
D-Bus <firstterm>message bus</firstterm>, specified in <xref
linkend="message-bus"/>. The message bus is a special application that
accepts connections from multiple other applications, and forwards
messages among them.
</para>
<para>
Uses of D-Bus include notification of system changes (notification of when
a camera is plugged in to a computer, or a new version of some software
has been installed), or desktop interoperability, for example a file
monitoring service or a configuration service.
</para>
<para>
D-Bus is designed for two specific use cases:
<itemizedlist>
<listitem>
<para>
A "system bus" for notifications from the system to user sessions,
and to allow the system to request input from user sessions.
</para>
</listitem>
<listitem>
<para>
A "session bus" used to implement desktop environments such as
GNOME and KDE.
</para>
</listitem>
</itemizedlist>
D-Bus is not intended to be a generic IPC system for any possible
application, and intentionally omits many features found in other
IPC systems for this reason.
</para>
<para>
At the same time, the bus daemons offer a number of features not found in
other IPC systems, such as single-owner "bus names" (similar to X
selections), on-demand startup of services, and security policies.
In many ways, these features are the primary motivation for developing
D-Bus; other systems would have sufficed if IPC were the only goal.
</para>
<para>
D-Bus may turn out to be useful in unanticipated applications, but future
versions of this spec and the reference implementation probably will not
incorporate features that interfere with the core use cases.
</para>
<para>
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in RFC 2119. However, the
document could use a serious audit to be sure it makes sense to do
so. Also, they are not capitalized.
</para>
<sect2 id="stability">
<title>Protocol and Specification Stability</title>
<para>
The D-Bus protocol is frozen (only compatible extensions are allowed) as
of November 8, 2006. However, this specification could still use a fair
bit of work to make interoperable reimplementation possible without
reference to the D-Bus reference implementation. Thus, this
specification is not marked 1.0. To mark it 1.0, we'd like to see
someone invest significant effort in clarifying the specification
language, and growing the specification to cover more aspects of the
reference implementation's behavior.
</para>
<para>
Until this work is complete, any attempt to reimplement D-Bus will
probably require looking at the reference implementation and/or asking
questions on the D-Bus mailing list about intended behavior.
Questions on the list are very welcome.
</para>
<para>
Nonetheless, this document should be a useful starting point and is
to our knowledge accurate, though incomplete.
</para>
</sect2>
</sect1>
<sect1 id="type-system">
<title>Type System</title>
<para>
D-Bus has a type system, in which values of various types can be
serialized into a sequence of bytes referred to as the
<firstterm>wire format</firstterm> in a standard way.
Converting a value from some other representation into the wire
format is called <firstterm>marshaling</firstterm> and converting
it back from the wire format is <firstterm>unmarshaling</firstterm>.
</para>
<sect2 id="message-protocol-signatures">
<title>Type Signatures</title>
<para>
The D-Bus protocol does not include type tags in the marshaled data; a
block of marshaled values must have a known <firstterm>type
signature</firstterm>. The type signature is made up of <firstterm>type
codes</firstterm>. A type code is an ASCII character representing the
type of a value. Because ASCII characters are used, the type signature
will always form a valid ASCII string. A simple string compare
determines whether two type signatures are equivalent.
</para>
<para>
As a simple example, the type code for 32-bit integer (<literal>INT32</literal>) is
the ASCII character 'i'. So the signature for a block of values
containing a single <literal>INT32</literal> would be:
<programlisting>
"i"
</programlisting>
A block of values containing two <literal>INT32</literal> would have this signature:
<programlisting>
"ii"
</programlisting>
</para>
<para>
All <firstterm>basic</firstterm> types work like
<literal>INT32</literal> in this example. To marshal and unmarshal
basic types, you simply read one value from the data
block corresponding to each type code in the signature.
In addition to basic types, there are four <firstterm>container</firstterm>
types: <literal>STRUCT</literal>, <literal>ARRAY</literal>, <literal>VARIANT</literal>,
and <literal>DICT_ENTRY</literal>.
</para>
<para>
<literal>STRUCT</literal> has a type code, ASCII character 'r', but this type
code does not appear in signatures. Instead, ASCII characters
'(' and ')' are used to mark the beginning and end of the struct.
So for example, a struct containing two integers would have this
signature:
<programlisting>
"(ii)"
</programlisting>
Structs can be nested, so for example a struct containing
an integer and another struct:
<programlisting>
"(i(ii))"
</programlisting>
The value block storing that struct would contain three integers; the
type signature allows you to distinguish "(i(ii))" from "((ii)i)" or
"(iii)" or "iii".
</para>
<para>
The <literal>STRUCT</literal> type code 'r' is not currently used in the D-Bus protocol,
but is useful in code that implements the protocol. This type code
is specified to allow such code to interoperate in non-protocol contexts.
</para>
<para>
Empty structures are not allowed; there must be at least one
type code between the parentheses.
</para>
<para>
<literal>ARRAY</literal> has ASCII character 'a' as type code. The array type code must be
followed by a <firstterm>single complete type</firstterm>. The single
complete type following the array is the type of each array element. So
the simple example is:
<programlisting>
"ai"
</programlisting>
which is an array of 32-bit integers. But an array can be of any type,
such as this array-of-struct-with-two-int32-fields:
<programlisting>
"a(ii)"
</programlisting>
Or this array of array of integer:
<programlisting>
"aai"
</programlisting>
</para>
<para>
The phrase <firstterm>single complete type</firstterm> deserves some
definition. A single complete type is a basic type code, a variant type code,
an array with its element type, or a struct with its fields.
So the following signatures are not single complete types:
<programlisting>
"aa"
</programlisting>
<programlisting>
"(ii"
</programlisting>
<programlisting>
"ii)"
</programlisting>
And the following signatures contain multiple complete types:
<programlisting>
"ii"
</programlisting>
<programlisting>
"aiai"
</programlisting>
<programlisting>
"(ii)(ii)"
</programlisting>
Note however that a single complete type may <emphasis>contain</emphasis>
multiple other single complete types.
</para>
<para>
<literal>VARIANT</literal> has ASCII character 'v' as its type code. A marshaled value of
type <literal>VARIANT</literal> will have the signature of a single complete type as part
of the <emphasis>value</emphasis>. This signature will be followed by a
marshaled value of that type.
</para>
<para>
A <literal>DICT_ENTRY</literal> works exactly like a struct, but rather
than parentheses it uses curly braces, and it has more restrictions.
The restrictions are: it occurs only as an array element type; it has
exactly two single complete types inside the curly braces; the first
single complete type (the "key") must be a basic type rather than a
container type. Implementations must not accept dict entries outside of
arrays, must not accept dict entries with zero, one, or more than two
fields, and must not accept dict entries with non-basic-typed keys. A
dict entry is always a key-value pair.
</para>
<para>
The first field in the <literal>DICT_ENTRY</literal> is always the key.
A message is considered corrupt if the same key occurs twice in the same
array of <literal>DICT_ENTRY</literal>. However, for performance reasons
implementations are not required to reject dicts with duplicate keys.
</para>
<para>
In most languages, an array of dict entry would be represented as a
map, hash table, or dict object.
</para>
<para>
The following table summarizes the D-Bus types.
<informaltable>
<tgroup cols="3">
<thead>
<row>
<entry>Conventional Name</entry>
<entry>Code</entry>
<entry>Description</entry>
</row>
</thead>
<tbody>
<row>
<entry><literal>INVALID</literal></entry>
<entry>0 (ASCII NUL)</entry>
<entry>Not a valid type code, used to terminate signatures</entry>
</row><row>
<entry><literal>BYTE</literal></entry>
<entry>121 (ASCII 'y')</entry>
<entry>8-bit unsigned integer</entry>
</row><row>
<entry><literal>BOOLEAN</literal></entry>
<entry>98 (ASCII 'b')</entry>
<entry>Boolean value, 0 is <literal>FALSE</literal> and 1 is <literal>TRUE</literal>. Everything else is invalid.</entry>
</row><row>
<entry><literal>INT16</literal></entry>
<entry>110 (ASCII 'n')</entry>
<entry>16-bit signed integer</entry>
</row><row>
<entry><literal>UINT16</literal></entry>
<entry>113 (ASCII 'q')</entry>
<entry>16-bit unsigned integer</entry>
</row><row>
<entry><literal>INT32</literal></entry>
<entry>105 (ASCII 'i')</entry>
<entry>32-bit signed integer</entry>
</row><row>
<entry><literal>UINT32</literal></entry>
<entry>117 (ASCII 'u')</entry>
<entry>32-bit unsigned integer</entry>
</row><row>
<entry><literal>INT64</literal></entry>
<entry>120 (ASCII 'x')</entry>
<entry>64-bit signed integer</entry>
</row><row>
<entry><literal>UINT64</literal></entry>
<entry>116 (ASCII 't')</entry>
<entry>64-bit unsigned integer</entry>
</row><row>
<entry><literal>DOUBLE</literal></entry>
<entry>100 (ASCII 'd')</entry>
<entry>IEEE 754 double</entry>
</row><row>
<entry><literal>STRING</literal></entry>
<entry>115 (ASCII 's')</entry>
<entry>UTF-8 string (<emphasis>must</emphasis> be valid UTF-8). Must be nul terminated and contain no other nul bytes.</entry>
</row><row>
<entry><literal>OBJECT_PATH</literal></entry>
<entry>111 (ASCII 'o')</entry>
<entry>Name of an object instance</entry>
</row><row>
<entry><literal>SIGNATURE</literal></entry>
<entry>103 (ASCII 'g')</entry>
<entry>A type signature</entry>
</row><row>
<entry><literal>ARRAY</literal></entry>
<entry>97 (ASCII 'a')</entry>
<entry>Array</entry>
</row><row>
<entry><literal>STRUCT</literal></entry>
<entry>114 (ASCII 'r'), 40 (ASCII '('), 41 (ASCII ')')</entry>
<entry>Struct; type code 114 'r' is reserved for use in
bindings and implementations to represent the general
concept of a struct, and must not appear in signatures
used on D-Bus.</entry>
</row><row>
<entry><literal>VARIANT</literal></entry>
<entry>118 (ASCII 'v') </entry>
<entry>Variant type (the type of the value is part of the value itself)</entry>
</row><row>
<entry><literal>DICT_ENTRY</literal></entry>
<entry>101 (ASCII 'e'), 123 (ASCII '{'), 125 (ASCII '}') </entry>
<entry>Entry in a dict or map (array of key-value pairs).
Type code 101 'e' is reserved for use in bindings and
implementations to represent the general concept of a
dict or dict-entry, and must not appear in signatures
used on D-Bus.</entry>
</row><row>
<entry><literal>UNIX_FD</literal></entry>
<entry>104 (ASCII 'h')</entry>
<entry>Unix file descriptor</entry>
</row>
<row>
<entry>(reserved)</entry>
<entry>109 (ASCII 'm')</entry>
<entry>Reserved for <ulink
url="https://bugs.freedesktop.org/show_bug.cgi?id=27857">a
'maybe' type compatible with the one in GVariant</ulink>,
and must not appear in signatures used on D-Bus until
specified here</entry>
</row>
<row>
<entry>(reserved)</entry>
<entry>42 (ASCII '*')</entry>
<entry>Reserved for use in bindings/implementations to
represent any <firstterm>single complete type</firstterm>,
and must not appear in signatures used on D-Bus.</entry>
</row>
<row>
<entry>(reserved)</entry>
<entry>63 (ASCII '?')</entry>
<entry>Reserved for use in bindings/implementations to
represent any <firstterm>basic type</firstterm>, and must
not appear in signatures used on D-Bus.</entry>
</row>
<row>
<entry>(reserved)</entry>
<entry>64 (ASCII '@'), 38 (ASCII '&amp;'),
94 (ASCII '^')</entry>
<entry>Reserved for internal use by bindings/implementations,
and must not appear in signatures used on D-Bus.
GVariant uses these type-codes to encode calling
conventions.</entry>
</row>
</tbody>
</tgroup>
</informaltable>
</para>
</sect2>
<sect2 id="message-protocol-marshaling">
<title>Marshaling (Wire Format)</title>
<para>
Given a type signature, a block of bytes can be converted into typed
values. This section describes the format of the block of bytes. Byte
order and alignment issues are handled uniformly for all D-Bus types.
</para>
<para>
A block of bytes has an associated byte order. The byte order
has to be discovered in some way; for D-Bus messages, the
byte order is part of the message header as described in
<xref linkend="message-protocol-messages"/>. For now, assume
that the byte order is known to be either little endian or big
endian.
</para>
<para>
Each value in a block of bytes is aligned "naturally," for example
4-byte values are aligned to a 4-byte boundary, and 8-byte values to an
8-byte boundary. To properly align a value, <firstterm>alignment
padding</firstterm> may be necessary. The alignment padding must always
be the minimum required padding to properly align the following value;
and it must always be made up of nul bytes. The alignment padding must
not be left uninitialized (it can't contain garbage), and more padding
than required must not be used.
</para>
<para>
Given all this, the types are marshaled on the wire as follows:
<informaltable>
<tgroup cols="3">
<thead>
<row>
<entry>Conventional Name</entry>
<entry>Encoding</entry>
<entry>Alignment</entry>
</row>
</thead>
<tbody>
<row>
<entry><literal>INVALID</literal></entry>
<entry>Not applicable; cannot be marshaled.</entry>
<entry>N/A</entry>
</row><row>
<entry><literal>BYTE</literal></entry>
<entry>A single 8-bit byte.</entry>
<entry>1</entry>
</row><row>
<entry><literal>BOOLEAN</literal></entry>
<entry>As for <literal>UINT32</literal>, but only 0 and 1 are valid values.</entry>
<entry>4</entry>
</row><row>
<entry><literal>INT16</literal></entry>
<entry>16-bit signed integer in the message's byte order.</entry>
<entry>2</entry>
</row><row>
<entry><literal>UINT16</literal></entry>
<entry>16-bit unsigned integer in the message's byte order.</entry>
<entry>2</entry>
</row><row>
<entry><literal>INT32</literal></entry>
<entry>32-bit signed integer in the message's byte order.</entry>
<entry>4</entry>
</row><row>
<entry><literal>UINT32</literal></entry>
<entry>32-bit unsigned integer in the message's byte order.</entry>
<entry>4</entry>
</row><row>
<entry><literal>INT64</literal></entry>
<entry>64-bit signed integer in the message's byte order.</entry>
<entry>8</entry>
</row><row>
<entry><literal>UINT64</literal></entry>
<entry>64-bit unsigned integer in the message's byte order.</entry>
<entry>8</entry>
</row><row>
<entry><literal>DOUBLE</literal></entry>
<entry>64-bit IEEE 754 double in the message's byte order.</entry>
<entry>8</entry>
</row><row>
<entry><literal>STRING</literal></entry>
<entry>A <literal>UINT32</literal> indicating the string's
length in bytes excluding its terminating nul, followed by
non-nul string data of the given length, followed by a terminating nul
byte.
</entry>
<entry>
4 (for the length)
</entry>
</row><row>
<entry><literal>OBJECT_PATH</literal></entry>
<entry>Exactly the same as <literal>STRING</literal> except the
content must be a valid object path (see below).
</entry>
<entry>
4 (for the length)
</entry>
</row><row>
<entry><literal>SIGNATURE</literal></entry>
<entry>The same as <literal>STRING</literal> except the length is a single
byte (thus signatures have a maximum length of 255)
and the content must be a valid signature (see below).
</entry>
<entry>
1
</entry>
</row><row>
<entry><literal>ARRAY</literal></entry>
<entry>
A <literal>UINT32</literal> giving the length of the array data in bytes, followed by
alignment padding to the alignment boundary of the array element type,
followed by each array element. The array length is from the
end of the alignment padding to the end of the last element,
i.e. it does not include the padding after the length,
or any padding after the last element.
Arrays have a maximum length defined to be 2 to the 26th power or
67108864. Implementations must not send or accept arrays exceeding this
length.
</entry>
<entry>
4 (for the length)
</entry>
</row><row>
<entry><literal>STRUCT</literal></entry>
<entry>
A struct must start on an 8-byte boundary regardless of the
type of the struct fields. The struct value consists of each
field marshaled in sequence starting from that 8-byte
alignment boundary.
</entry>
<entry>
8
</entry>
</row><row>
<entry><literal>VARIANT</literal></entry>
<entry>
A variant type has a marshaled
<literal>SIGNATURE</literal> followed by a marshaled
value with the type given in the signature. Unlike
a message signature, the variant signature can
contain only a single complete type. So "i", "ai"
or "(ii)" is OK, but "ii" is not. Use of variants may not
cause a total message depth to be larger than 64, including
other container types such as structures.
</entry>
<entry>
1 (alignment of the signature)
</entry>
</row><row>
<entry><literal>DICT_ENTRY</literal></entry>
<entry>
Identical to STRUCT.
</entry>
<entry>
8
</entry>
</row><row>
<entry><literal>UNIX_FD</literal></entry>
<entry>32-bit unsigned integer in the message's byte
order. The actual file descriptors need to be
transferred out-of-band via some platform specific
mechanism. On the wire, values of this type store the index to the
file descriptor in the array of file descriptors that
accompany the message.</entry>
<entry>4</entry>
</row>
</tbody>
</tgroup>
</informaltable>
</para>
<sect3 id="message-protocol-marshaling-object-path">
<title>Valid Object Paths</title>
<para>
An object path is a name used to refer to an object instance.
Conceptually, each participant in a D-Bus message exchange may have
any number of object instances (think of C++ or Java objects) and each
such instance will have a path. Like a filesystem, the object
instances in an application form a hierarchical tree.
</para>
<para>
The following rules define a valid object path. Implementations must
not send or accept messages with invalid object paths.
<itemizedlist>
<listitem>
<para>
The path may be of any length.
</para>
</listitem>
<listitem>
<para>
The path must begin with an ASCII '/' (integer 47) character,
and must consist of elements separated by slash characters.
</para>
</listitem>
<listitem>
<para>
Each element must only contain the ASCII characters
"[A-Z][a-z][0-9]_"
</para>
</listitem>
<listitem>
<para>
No element may be the empty string.
</para>
</listitem>
<listitem>
<para>
Multiple '/' characters cannot occur in sequence.
</para>
</listitem>
<listitem>
<para>
A trailing '/' character is not allowed unless the
path is the root path (a single '/' character).
</para>
</listitem>
</itemizedlist>
</para>
<para>
Object paths are often namespaced by starting with a reversed
domain name and containing an interface version number, in the
same way as
<link linkend="message-protocol-names-interface">interface
names</link> and
<link linkend="message-protocol-names-bus">well-known
bus names</link>.
This makes it possible to implement more than one service, or
more than one version of a service, in the same process,
even if the services share a connection but cannot otherwise
co-operate (for instance, if they are implemented by different
plugins).
</para>
<para>
For instance, if the owner of <literal>example.com</literal> is
developing a D-Bus API for a music player, they might use the
hierarchy of object paths that start with
<literal>/com/example/MusicPlayer1</literal> for its objects.
</para>
</sect3>
<sect3 id="message-protocol-marshaling-signature">
<title>Valid Signatures</title>
<para>
An implementation must not send or accept invalid signatures.
Valid signatures will conform to the following rules:
<itemizedlist>
<listitem>
<para>
The signature ends with a nul byte.
</para>
</listitem>
<listitem>
<para>
The signature is a list of single complete types.
Arrays must have element types, and structs must
have both open and close parentheses.
</para>
</listitem>
<listitem>
<para>
Only type codes and open and close parentheses are
allowed in the signature. The <literal>STRUCT</literal> type code
is not allowed in signatures, because parentheses
are used instead.
</para>
</listitem>
<listitem>
<para>
The maximum depth of container type nesting is 32 array type
codes and 32 open parentheses. This implies that the maximum
total depth of recursion is 64, for an "array of array of array
of ... struct of struct of struct of ..." where there are 32
array and 32 struct.
</para>
</listitem>
<listitem>
<para>
The maximum length of a signature is 255.
</para>
</listitem>
<listitem>
<para>
Signatures must be nul-terminated.
</para>
</listitem>
</itemizedlist>
</para>
</sect3>
</sect2>
</sect1>
<sect1 id="message-protocol">
<title>Message Protocol</title>
<para>
A <firstterm>message</firstterm> consists of a
<firstterm>header</firstterm> and a <firstterm>body</firstterm>. If you
think of a message as a package, the header is the address, and the body
contains the package contents. The message delivery system uses the header
information to figure out where to send the message and how to interpret
it; the recipient interprets the body of the message.
</para>
<para>
The body of the message is made up of zero or more
<firstterm>arguments</firstterm>, which are typed values, such as an
integer or a byte array.
</para>
<para>
Both header and body use the D-Bus <link linkend="type-system">type
system</link> and format for serializing data.
</para>
<sect2 id="message-protocol-messages">
<title>Message Format</title>
<para>
A message consists of a header and a body. The header is a block of
values with a fixed signature and meaning. The body is a separate block
of values, with a signature specified in the header.
</para>
<para>
The length of the header must be a multiple of 8, allowing the body to
begin on an 8-byte boundary when storing the entire message in a single
buffer. If the header does not naturally end on an 8-byte boundary
up to 7 bytes of nul-initialized alignment padding must be added.
</para>
<para>
The message body need not end on an 8-byte boundary.
</para>
<para>
The maximum length of a message, including header, header alignment padding,
and body is 2 to the 27th power or 134217728. Implementations must not
send or accept messages exceeding this size.
</para>
<para>
The signature of the header is:
<programlisting>
"yyyyuua(yv)"
</programlisting>
Written out more readably, this is:
<programlisting>
BYTE, BYTE, BYTE, BYTE, UINT32, UINT32, ARRAY of STRUCT of (BYTE,VARIANT)
</programlisting>
</para>
<para>
These values have the following meanings:
<informaltable>
<tgroup cols="2">
<thead>
<row>
<entry>Value</entry>
<entry>Description</entry>
</row>
</thead>
<tbody>
<row>
<entry>1st <literal>BYTE</literal></entry>
<entry>Endianness flag; ASCII 'l' for little-endian
or ASCII 'B' for big-endian. Both header and body are
in this endianness.</entry>
</row>
<row>
<entry>2nd <literal>BYTE</literal></entry>
<entry><firstterm>Message type</firstterm>. Unknown types must be ignored.
Currently-defined types are described below.
</entry>
</row>
<row>
<entry>3rd <literal>BYTE</literal></entry>
<entry>Bitwise OR of flags. Unknown flags
must be ignored. Currently-defined flags are described below.
</entry>
</row>
<row>
<entry>4th <literal>BYTE</literal></entry>
<entry>Major protocol version of the sending application. If
the major protocol version of the receiving application does not
match, the applications will not be able to communicate and the
D-Bus connection must be disconnected. The major protocol
version for this version of the specification is 1.
</entry>
</row>
<row>
<entry>1st <literal>UINT32</literal></entry>
<entry>Length in bytes of the message body, starting
from the end of the header. The header ends after
its alignment padding to an 8-boundary.
</entry>
</row>
<row>
<entry>2nd <literal>UINT32</literal></entry>
<entry>The serial of this message, used as a cookie
by the sender to identify the reply corresponding
to this request. This must not be zero.
</entry>
</row>
<row>
<entry><literal>ARRAY</literal> of <literal>STRUCT</literal> of (<literal>BYTE</literal>,<literal>VARIANT</literal>)</entry>
<entry>An array of zero or more <firstterm>header
fields</firstterm> where the byte is the field code, and the
variant is the field value. The message type determines
which fields are required.
</entry>
</row>
</tbody>
</tgroup>
</informaltable>
</para>
<para>
<firstterm>Message types</firstterm> that can appear in the second byte
of the header are:
<informaltable>
<tgroup cols="3">
<thead>
<row>
<entry>Conventional name</entry>
<entry>Decimal value</entry>
<entry>Description</entry>
</row>
</thead>
<tbody>
<row>
<entry><literal>INVALID</literal></entry>
<entry>0</entry>
<entry>This is an invalid type.</entry>
</row>
<row>
<entry><literal>METHOD_CALL</literal></entry>
<entry>1</entry>
<entry>Method call.</entry>
</row>
<row>
<entry><literal>METHOD_RETURN</literal></entry>
<entry>2</entry>
<entry>Method reply with returned data.</entry>
</row>
<row>
<entry><literal>ERROR</literal></entry>
<entry>3</entry>
<entry>Error reply. If the first argument exists and is a
string, it is an error message.</entry>
</row>
<row>
<entry><literal>SIGNAL</literal></entry>
<entry>4</entry>
<entry>Signal emission.</entry>
</row>
</tbody>
</tgroup>
</informaltable>
</para>
<para>
Flags that can appear in the third byte of the header:
<informaltable>
<tgroup cols="3">
<thead>
<row>
<entry>Conventional name</entry>
<entry>Hex value</entry>
<entry>Description</entry>
</row>
</thead>
<tbody>
<row>
<entry><literal>NO_REPLY_EXPECTED</literal></entry>
<entry>0x1</entry>
<entry>This message does not expect method return replies or
error replies; the reply can be omitted as an
optimization. However, it is compliant with this specification
to return the reply despite this flag and the only harm
from doing so is extra network traffic.
</entry>
</row>
<row>
<entry><literal>NO_AUTO_START</literal></entry>
<entry>0x2</entry>
<entry>The bus must not launch an owner
for the destination name in response to this message.
</entry>
</row>
</tbody>
</tgroup>
</informaltable>
</para>
<sect3 id="message-protocol-header-fields">
<title>Header Fields</title>
<para>
The array at the end of the header contains <firstterm>header
fields</firstterm>, where each field is a 1-byte field code followed
by a field value. A header must contain the required header fields for
its message type, and zero or more of any optional header
fields. Future versions of this protocol specification may add new
fields. Implementations must ignore fields they do not
understand. Implementations must not invent their own header fields;
only changes to this specification may introduce new header fields.
</para>
<para>
Again, if an implementation sees a header field code that it does not
expect, it must ignore that field, as it will be part of a new
(but compatible) version of this specification. This also applies
to known header fields appearing in unexpected messages, for
example: if a signal has a reply serial it must be ignored
even though it has no meaning as of this version of the spec.
</para>
<para>
However, implementations must not send or accept known header fields
with the wrong type stored in the field value. So for example a
message with an <literal>INTERFACE</literal> field of type
<literal>UINT32</literal> would be considered corrupt.
</para>
<para>
Here are the currently-defined header fields:
<informaltable>
<tgroup cols="5">
<thead>
<row>
<entry>Conventional Name</entry>
<entry>Decimal Code</entry>
<entry>Type</entry>
<entry>Required In</entry>
<entry>Description</entry>
</row>
</thead>
<tbody>
<row>
<entry><literal>INVALID</literal></entry>
<entry>0</entry>
<entry>N/A</entry>
<entry>not allowed</entry>
<entry>Not a valid field name (error if it appears in a message)</entry>
</row>
<row>
<entry><literal>PATH</literal></entry>
<entry>1</entry>
<entry><literal>OBJECT_PATH</literal></entry>
<entry><literal>METHOD_CALL</literal>, <literal>SIGNAL</literal></entry>
<entry>The object to send a call to,
or the object a signal is emitted from.
The special path
<literal>/org/freedesktop/DBus/Local</literal> is reserved;
implementations should not send messages with this path,
and the reference implementation of the bus daemon will
disconnect any application that attempts to do so.
</entry>
</row>
<row>
<entry><literal>INTERFACE</literal></entry>
<entry>2</entry>
<entry><literal>STRING</literal></entry>
<entry><literal>SIGNAL</literal></entry>
<entry>
The interface to invoke a method call on, or
that a signal is emitted from. Optional for
method calls, required for signals.
The special interface
<literal>org.freedesktop.DBus.Local</literal> is reserved;
implementations should not send messages with this
interface, and the reference implementation of the bus
daemon will disconnect any application that attempts to
do so.
</entry>
</row>
<row>
<entry><literal>MEMBER</literal></entry>
<entry>3</entry>
<entry><literal>STRING</literal></entry>
<entry><literal>METHOD_CALL</literal>, <literal>SIGNAL</literal></entry>
<entry>The member, either the method name or signal name.</entry>
</row>
<row>
<entry><literal>ERROR_NAME</literal></entry>
<entry>4</entry>
<entry><literal>STRING</literal></entry>
<entry><literal>ERROR</literal></entry>
<entry>The name of the error that occurred, for errors</entry>
</row>
<row>
<entry><literal>REPLY_SERIAL</literal></entry>
<entry>5</entry>
<entry><literal>UINT32</literal></entry>
<entry><literal>ERROR</literal>, <literal>METHOD_RETURN</literal></entry>
<entry>The serial number of the message this message is a reply
to. (The serial number is the second <literal>UINT32</literal> in the header.)</entry>
</row>
<row>
<entry><literal>DESTINATION</literal></entry>
<entry>6</entry>
<entry><literal>STRING</literal></entry>
<entry>optional</entry>
<entry>The name of the connection this message is intended for.
Only used in combination with the message bus, see
<xref linkend="message-bus"/>.</entry>
</row>
<row>
<entry><literal>SENDER</literal></entry>
<entry>7</entry>
<entry><literal>STRING</literal></entry>
<entry>optional</entry>
<entry>Unique name of the sending connection.
The message bus fills in this field so it is reliable; the field is
only meaningful in combination with the message bus.</entry>
</row>
<row>
<entry><literal>SIGNATURE</literal></entry>
<entry>8</entry>
<entry><literal>SIGNATURE</literal></entry>
<entry>optional</entry>
<entry>The signature of the message body.
If omitted, it is assumed to be the
empty signature "" (i.e. the body must be 0-length).</entry>
</row>
<row>
<entry><literal>UNIX_FDS</literal></entry>
<entry>9</entry>
<entry><literal>UINT32</literal></entry>
<entry>optional</entry>
<entry>The number of Unix file descriptors that
accompany the message. If omitted, it is assumed
that no Unix file descriptors accompany the
message. The actual file descriptors need to be
transferred via platform specific mechanism
out-of-band. They must be sent at the same time as
part of the message itself. They may not be sent
before the first byte of the message itself is
transferred or after the last byte of the message
itself.</entry>
</row>
</tbody>
</tgroup>
</informaltable>
</para>
</sect3>
</sect2>
<sect2 id="message-protocol-names">
<title>Valid Names</title>
<para>
The various names in D-Bus messages have some restrictions.
</para>
<para>
There is a <firstterm>maximum name length</firstterm>
of 255 which applies to bus names, interfaces, and members.
</para>
<sect3 id="message-protocol-names-interface">
<title>Interface names</title>
<para>
Interfaces have names with type <literal>STRING</literal>, meaning that
they must be valid UTF-8. However, there are also some
additional restrictions that apply to interface names
specifically:
<itemizedlist>
<listitem><para>Interface names are composed of 1 or more elements separated by
a period ('.') character. All elements must contain at least
one character.
</para>
</listitem>
<listitem><para>Each element must only contain the ASCII characters
"[A-Z][a-z][0-9]_" and must not begin with a digit.
</para>
</listitem>
<listitem><para>Interface names must contain at least one '.' (period)
character (and thus at least two elements).
</para></listitem>
<listitem><para>Interface names must not begin with a '.' (period) character.</para></listitem>
<listitem><para>Interface names must not exceed the maximum name length.</para></listitem>
</itemizedlist>
</para>
<para>
Interface names should start with the reversed DNS domain name of
the author of the interface (in lower-case), like interface names
in Java. It is conventional for the rest of the interface name
to consist of words run together, with initial capital letters
on all words ("CamelCase"). Several levels of hierarchy can be used.
It is also a good idea to include the major version of the interface
in the name, and increment it if incompatible changes are made;
this way, a single object can implement several versions of an
interface in parallel, if necessary.
</para>
<para>
For instance, if the owner of <literal>example.com</literal> is
developing a D-Bus API for a music player, they might define
interfaces called <literal>com.example.MusicPlayer1</literal>,
<literal>com.example.MusicPlayer1.Track</literal> and
<literal>com.example.MusicPlayer1.Seekable</literal>.
</para>
<para>
D-Bus does not distinguish between the concepts that would be
called classes and interfaces in Java: either can be identified on
D-Bus by an interface name.
</para>
</sect3>
<sect3 id="message-protocol-names-bus">
<title>Bus names</title>
<para>
Connections have one or more bus names associated with them.
A connection has exactly one bus name that is a <firstterm>unique
connection name</firstterm>. The unique connection name remains
with the connection for its entire lifetime.
A bus name is of type <literal>STRING</literal>,
meaning that it must be valid UTF-8. However, there are also
some additional restrictions that apply to bus names
specifically:
<itemizedlist>
<listitem><para>Bus names that start with a colon (':')
character are unique connection names. Other bus names
are called <firstterm>well-known bus names</firstterm>.
</para>
</listitem>
<listitem><para>Bus names are composed of 1 or more elements separated by
a period ('.') character. All elements must contain at least
one character.
</para>
</listitem>
<listitem><para>Each element must only contain the ASCII characters
"[A-Z][a-z][0-9]_-". Only elements that are part of a unique
connection name may begin with a digit, elements in
other bus names must not begin with a digit.
</para>
</listitem>
<listitem><para>Bus names must contain at least one '.' (period)
character (and thus at least two elements).
</para></listitem>
<listitem><para>Bus names must not begin with a '.' (period) character.</para></listitem>
<listitem><para>Bus names must not exceed the maximum name length.</para></listitem>
</itemizedlist>
</para>
<para>
Note that the hyphen ('-') character is allowed in bus names but
not in interface names.
</para>
<para>
Like <link linkend="message-protocol-names-interface">interface
names</link>, well-known bus names should start with the
reversed DNS domain name of the author of the interface (in
lower-case), and it is conventional for the rest of the well-known
bus name to consist of words run together, with initial
capital letters. As with interface names, including a version
number in well-known bus names is a good idea; it's possible to
have the well-known bus name for more than one version
simultaneously if backwards compatibility is required.
</para>
<para>
If a well-known bus name implies the presence of a "main" interface,
that "main" interface is often given the same name as
the well-known bus name, and situated at the corresponding object
path. For instance, if the owner of <literal>example.com</literal>
is developing a D-Bus API for a music player, they might define
that any application that takes the well-known name
<literal>com.example.MusicPlayer1</literal> should have an object
at the object path <literal>/com/example/MusicPlayer1</literal>
which implements the interface
<literal>com.example.MusicPlayer1</literal>.
</para>
</sect3>
<sect3 id="message-protocol-names-member">
<title>Member names</title>
<para>
Member (i.e. method or signal) names:
<itemizedlist>
<listitem><para>Must only contain the ASCII characters
"[A-Z][a-z][0-9]_" and may not begin with a
digit.</para></listitem>
<listitem><para>Must not contain the '.' (period) character.</para></listitem>
<listitem><para>Must not exceed the maximum name length.</para></listitem>
<listitem><para>Must be at least 1 byte in length.</para></listitem>
</itemizedlist>
</para>
<para>
It is conventional for member names on D-Bus to consist of
capitalized words with no punctuation ("camel-case").
Method names should usually be verbs, such as
<literal>GetItems</literal>, and signal names should usually be
a description of an event, such as <literal>ItemsChanged</literal>.
</para>
</sect3>
<sect3 id="message-protocol-names-error">
<title>Error names</title>
<para>
Error names have the same restrictions as interface names.
</para>
<para>
Error names have the same naming conventions as interface
names, and often contain <literal>.Error.</literal>; for instance,
the owner of <literal>example.com</literal> might define the
errors <literal>com.example.MusicPlayer.Error.FileNotFound</literal>
and <literal>com.example.MusicPlayer.Error.OutOfMemory</literal>.
The errors defined by D-Bus itself, such as
<literal>org.freedesktop.DBus.Error.Failed</literal>, follow a
similar pattern.
</para>
</sect3>
</sect2>
<sect2 id="message-protocol-types">
<title>Message Types</title>
<para>
Each of the message types (<literal>METHOD_CALL</literal>, <literal>METHOD_RETURN</literal>, <literal>ERROR</literal>, and
<literal>SIGNAL</literal>) has its own expected usage conventions and header fields.
This section describes these conventions.
</para>
<sect3 id="message-protocol-types-method">
<title>Method Calls</title>
<para>
Some messages invoke an operation on a remote object. These are
called method call messages and have the type tag <literal>METHOD_CALL</literal>. Such
messages map naturally to methods on objects in a typical program.
</para>
<para>
A method call message is required to have a <literal>MEMBER</literal> header field
indicating the name of the method. Optionally, the message has an
<literal>INTERFACE</literal> field giving the interface the method is a part of. In the
absence of an <literal>INTERFACE</literal> field, if two interfaces on the same object have
a method with the same name, it is undefined which of the two methods
will be invoked. Implementations may also choose to return an error in
this ambiguous case. However, if a method name is unique
implementations must not require an interface field.
</para>
<para>
Method call messages also include a <literal>PATH</literal> field
indicating the object to invoke the method on. If the call is passing
through a message bus, the message will also have a
<literal>DESTINATION</literal> field giving the name of the connection
to receive the message.
</para>
<para>
When an application handles a method call message, it is required to
return a reply. The reply is identified by a <literal>REPLY_SERIAL</literal> header field
indicating the serial number of the <literal>METHOD_CALL</literal> being replied to. The
reply can have one of two types; either <literal>METHOD_RETURN</literal> or <literal>ERROR</literal>.
</para>
<para>
If the reply has type <literal>METHOD_RETURN</literal>, the arguments to the reply message
are the return value(s) or "out parameters" of the method call.
If the reply has type <literal>ERROR</literal>, then an "exception" has been thrown,
and the call fails; no return value will be provided. It makes
no sense to send multiple replies to the same method call.
</para>
<para>
Even if a method call has no return values, a <literal>METHOD_RETURN</literal>
reply is required, so the caller will know the method
was successfully processed.
</para>
<para>
The <literal>METHOD_RETURN</literal> or <literal>ERROR</literal> reply message must have the <literal>REPLY_SERIAL</literal>
header field.
</para>
<para>
If a <literal>METHOD_CALL</literal> message has the flag <literal>NO_REPLY_EXPECTED</literal>,
then as an optimization the application receiving the method
call may choose to omit the reply message (regardless of
whether the reply would have been <literal>METHOD_RETURN</literal> or <literal>ERROR</literal>).
However, it is also acceptable to ignore the <literal>NO_REPLY_EXPECTED</literal>
flag and reply anyway.
</para>
<para>
Unless a message has the flag <literal>NO_AUTO_START</literal>, if the
destination name does not exist then a program to own the destination
name will be started before the message is delivered. The message
will be held until the new program is successfully started or has
failed to start; in case of failure, an error will be returned. This
flag is only relevant in the context of a message bus, it is ignored
during one-to-one communication with no intermediate bus.
</para>
<sect4 id="message-protocol-types-method-apis">
<title>Mapping method calls to native APIs</title>
<para>
APIs for D-Bus may map method calls to a method call in a specific
programming language, such as C++, or may map a method call written
in an IDL to a D-Bus message.
</para>
<para>
In APIs of this nature, arguments to a method are often termed "in"
(which implies sent in the <literal>METHOD_CALL</literal>), or "out" (which implies
returned in the <literal>METHOD_RETURN</literal>). Some APIs such as CORBA also have
"inout" arguments, which are both sent and received, i.e. the caller
passes in a value which is modified. Mapped to D-Bus, an "inout"
argument is equivalent to an "in" argument, followed by an "out"
argument. You can't pass things "by reference" over the wire, so
"inout" is purely an illusion of the in-process API.
</para>
<para>
Given a method with zero or one return values, followed by zero or more
arguments, where each argument may be "in", "out", or "inout", the
caller constructs a message by appending each "in" or "inout" argument,
in order. "out" arguments are not represented in the caller's message.
</para>
<para>
The recipient constructs a reply by appending first the return value
if any, then each "out" or "inout" argument, in order.
"in" arguments are not represented in the reply message.
</para>
<para>
Error replies are normally mapped to exceptions in languages that have
exceptions.
</para>
<para>
In converting from native APIs to D-Bus, it is perhaps nice to
map D-Bus naming conventions ("FooBar") to native conventions
such as "fooBar" or "foo_bar" automatically. This is OK
as long as you can say that the native API is one that
was specifically written for D-Bus. It makes the most sense
when writing object implementations that will be exported
over the bus. Object proxies used to invoke remote D-Bus
objects probably need the ability to call any D-Bus method,
and thus a magic name mapping like this could be a problem.
</para>
<para>
This specification doesn't require anything of native API bindings;
the preceding is only a suggested convention for consistency
among bindings.
</para>
</sect4>
</sect3>
<sect3 id="message-protocol-types-signal">
<title>Signal Emission</title>
<para>
Unlike method calls, signal emissions have no replies.
A signal emission is simply a single message of type <literal>SIGNAL</literal>.
It must have three header fields: <literal>PATH</literal> giving the object
the signal was emitted from, plus <literal>INTERFACE</literal> and <literal>MEMBER</literal> giving
the fully-qualified name of the signal. The <literal>INTERFACE</literal> header is required
for signals, though it is optional for method calls.
</para>
</sect3>
<sect3 id="message-protocol-types-errors">
<title>Errors</title>
<para>
Messages of type <literal>ERROR</literal> are most commonly replies
to a <literal>METHOD_CALL</literal>, but may be returned in reply
to any kind of message. The message bus for example
will return an <literal>ERROR</literal> in reply to a signal emission if
the bus does not have enough memory to send the signal.
</para>
<para>
An <literal>ERROR</literal> may have any arguments, but if the first
argument is a <literal>STRING</literal>, it must be an error message.
The error message may be logged or shown to the user
in some way.
</para>
</sect3>
<sect3 id="message-protocol-types-notation">
<title>Notation in this document</title>
<para>
This document uses a simple pseudo-IDL to describe particular method
calls and signals. Here is an example of a method call:
<programlisting>
org.freedesktop.DBus.StartServiceByName (in STRING name, in UINT32 flags,
out UINT32 resultcode)
</programlisting>
This means <literal>INTERFACE</literal> = org.freedesktop.DBus, <literal>MEMBER</literal> = StartServiceByName,
<literal>METHOD_CALL</literal> arguments are <literal>STRING</literal> and <literal>UINT32</literal>, <literal>METHOD_RETURN</literal> argument
is <literal>UINT32</literal>. Remember that the <literal>MEMBER</literal> field can't contain any '.' (period)
characters so it's known that the last part of the name in
the "IDL" is the member name.
</para>
<para>
In C++ that might end up looking like this:
<programlisting>
unsigned int org::freedesktop::DBus::StartServiceByName (const char *name,
unsigned int flags);
</programlisting>
or equally valid, the return value could be done as an argument:
<programlisting>
void org::freedesktop::DBus::StartServiceByName (const char *name,
unsigned int flags,
unsigned int *resultcode);
</programlisting>
It's really up to the API designer how they want to make
this look. You could design an API where the namespace wasn't used
in C++, using STL or Qt, using varargs, or whatever you wanted.
</para>
<para>
Signals are written as follows:
<programlisting>
org.freedesktop.DBus.NameLost (STRING name)
</programlisting>
Signals don't specify "in" vs. "out" because only
a single direction is possible.
</para>
<para>
It isn't especially encouraged to use this lame pseudo-IDL in actual
API implementations; you might use the native notation for the
language you're using, or you might use COM or CORBA IDL, for example.
</para>
</sect3>
</sect2>
<sect2 id="message-protocol-handling-invalid">
<title>Invalid Protocol and Spec Extensions</title>
<para>
For security reasons, the D-Bus protocol should be strictly parsed and
validated, with the exception of defined extension points. Any invalid
protocol or spec violations should result in immediately dropping the
connection without notice to the other end. Exceptions should be
carefully considered, e.g. an exception may be warranted for a
well-understood idiosyncrasy of a widely-deployed implementation. In
cases where the other end of a connection is 100% trusted and known to
be friendly, skipping validation for performance reasons could also make
sense in certain cases.
</para>
<para>
Generally speaking violations of the "must" requirements in this spec
should be considered possible attempts to exploit security, and violations
of the "should" suggestions should be considered legitimate (though perhaps
they should generate an error in some cases).
</para>
<para>
The following extension points are built in to D-Bus on purpose and must
not be treated as invalid protocol. The extension points are intended
for use by future versions of this spec, they are not intended for third
parties. At the moment, the only way a third party could extend D-Bus
without breaking interoperability would be to introduce a way to negotiate new
feature support as part of the auth protocol, using EXTENSION_-prefixed
commands. There is not yet a standard way to negotiate features.
<itemizedlist>
<listitem>
<para>
In the authentication protocol (see <xref linkend="auth-protocol"/>) unknown
commands result in an ERROR rather than a disconnect. This enables
future extensions to the protocol. Commands starting with EXTENSION_ are
reserved for third parties.
</para>
</listitem>
<listitem>
<para>
The authentication protocol supports pluggable auth mechanisms.
</para>
</listitem>
<listitem>
<para>
The address format (see <xref linkend="addresses"/>) supports new
kinds of transport.
</para>
</listitem>
<listitem>
<para>
Messages with an unknown type (something other than
<literal>METHOD_CALL</literal>, <literal>METHOD_RETURN</literal>,
<literal>ERROR</literal>, <literal>SIGNAL</literal>) are ignored.
Unknown-type messages must still be well-formed in the same way
as the known messages, however. They still have the normal
header and body.
</para>
</listitem>
<listitem>
<para>
Header fields with an unknown or unexpected field code must be ignored,
though again they must still be well-formed.
</para>
</listitem>
<listitem>
<para>
New standard interfaces (with new methods and signals) can of course be added.
</para>
</listitem>
</itemizedlist>
</para>
</sect2>
</sect1>
<sect1 id="auth-protocol">
<title>Authentication Protocol</title>
<para>
Before the flow of messages begins, two applications must
authenticate. A simple plain-text protocol is used for
authentication; this protocol is a SASL profile, and maps fairly
directly from the SASL specification. The message encoding is
NOT used here, only plain text messages.
</para>
<para>
In examples, "C:" and "S:" indicate lines sent by the client and
server respectively.
</para>
<sect2 id="auth-protocol-overview">
<title>Protocol Overview</title>
<para>
The protocol is a line-based protocol, where each line ends with
\r\n. Each line begins with an all-caps ASCII command name containing
only the character range [A-Z_], a space, then any arguments for the
command, then the \r\n ending the line. The protocol is
case-sensitive. All bytes must be in the ASCII character set.
Commands from the client to the server are as follows:
<itemizedlist>
<listitem><para>AUTH [mechanism] [initial-response]</para></listitem>
<listitem><para>CANCEL</para></listitem>
<listitem><para>BEGIN</para></listitem>
<listitem><para>DATA &lt;data in hex encoding&gt;</para></listitem>
<listitem><para>ERROR [human-readable error explanation]</para></listitem>
<listitem><para>NEGOTIATE_UNIX_FD</para></listitem>
</itemizedlist>
From server to client are as follows:
<itemizedlist>
<listitem><para>REJECTED &lt;space-separated list of mechanism names&gt;</para></listitem>
<listitem><para>OK &lt;GUID in hex&gt;</para></listitem>
<listitem><para>DATA &lt;data in hex encoding&gt;</para></listitem>
<listitem><para>ERROR</para></listitem>
<listitem><para>AGREE_UNIX_FD</para></listitem>
</itemizedlist>
</para>
<para>
Unofficial extensions to the command set must begin with the letters
"EXTENSION_", to avoid conflicts with future official commands.
For example, "EXTENSION_COM_MYDOMAIN_DO_STUFF".
</para>
</sect2>
<sect2 id="auth-nul-byte">
<title>Special credentials-passing nul byte</title>
<para>
Immediately after connecting to the server, the client must send a
single nul byte. This byte may be accompanied by credentials
information on some operating systems that use sendmsg() with
SCM_CREDS or SCM_CREDENTIALS to pass credentials over UNIX domain
sockets. However, the nul byte must be sent even on other kinds of
socket, and even on operating systems that do not require a byte to be
sent in order to transmit credentials. The text protocol described in
this document begins after the single nul byte. If the first byte
received from the client is not a nul byte, the server may disconnect
that client.
</para>
<para>
A nul byte in any context other than the initial byte is an error;
the protocol is ASCII-only.
</para>
<para>
The credentials sent along with the nul byte may be used with the
SASL mechanism EXTERNAL.
</para>
</sect2>
<sect2 id="auth-command-auth">
<title>AUTH command</title>
<para>
If an AUTH command has no arguments, it is a request to list
available mechanisms. The server must respond with a REJECTED
command listing the mechanisms it understands, or with an error.
</para>
<para>
If an AUTH command specifies a mechanism, and the server supports
said mechanism, the server should begin exchanging SASL
challenge-response data with the client using DATA commands.
</para>
<para>
If the server does not support the mechanism given in the AUTH
command, it must send either a REJECTED command listing the mechanisms
it does support, or an error.
</para>
<para>
If the [initial-response] argument is provided, it is intended for use
with mechanisms that have no initial challenge (or an empty initial
challenge), as if it were the argument to an initial DATA command. If
the selected mechanism has an initial challenge and [initial-response]
was provided, the server should reject authentication by sending
REJECTED.
</para>
<para>
If authentication succeeds after exchanging DATA commands,
an OK command must be sent to the client.
</para>
<para>
The first octet received by the server after the \r\n of the BEGIN
command from the client must be the first octet of the
authenticated/encrypted stream of D-Bus messages.
</para>
<para>
If BEGIN is received by the server, the first octet received
by the client after the \r\n of the OK command must be the
first octet of the authenticated/encrypted stream of D-Bus
messages.
</para>
</sect2>
<sect2 id="auth-command-cancel">
<title>CANCEL Command</title>
<para>
At any time up to sending the BEGIN command, the client may send a
CANCEL command. On receiving the CANCEL command, the server must
send a REJECTED command and abort the current authentication
exchange.
</para>
</sect2>
<sect2 id="auth-command-data">
<title>DATA Command</title>
<para>
The DATA command may come from either client or server, and simply
contains a hex-encoded block of data to be interpreted
according to the SASL mechanism in use.
</para>
<para>
Some SASL mechanisms support sending an "empty string";
FIXME we need some way to do this.
</para>
</sect2>
<sect2 id="auth-command-begin">
<title>BEGIN Command</title>
<para>
The BEGIN command acknowledges that the client has received an
OK command from the server, and that the stream of messages
is about to begin.
</para>
<para>
The first octet received by the server after the \r\n of the BEGIN
command from the client must be the first octet of the
authenticated/encrypted stream of D-Bus messages.
</para>
</sect2>
<sect2 id="auth-command-rejected">
<title>REJECTED Command</title>
<para>
The REJECTED command indicates that the current authentication
exchange has failed, and further exchange of DATA is inappropriate.
The client would normally try another mechanism, or try providing
different responses to challenges.
</para><para>
Optionally, the REJECTED command has a space-separated list of
available auth mechanisms as arguments. If a server ever provides
a list of supported mechanisms, it must provide the same list
each time it sends a REJECTED message. Clients are free to
ignore all lists received after the first.
</para>
</sect2>
<sect2 id="auth-command-ok">
<title>OK Command</title>
<para>
The OK command indicates that the client has been
authenticated. The client may now proceed with negotiating
Unix file descriptor passing. To do that it shall send
NEGOTIATE_UNIX_FD to the server.
</para>
<para>
Otherwise, the client must respond to the OK command by
sending a BEGIN command, followed by its stream of messages,
or by disconnecting. The server must not accept additional
commands using this protocol after the BEGIN command has been
received. Further communication will be a stream of D-Bus
messages (optionally encrypted, as negotiated) rather than
this protocol.
</para>
<para>
If a client sends BEGIN the first octet received by the client
after the \r\n of the OK command must be the first octet of
the authenticated/encrypted stream of D-Bus messages.
</para>
<para>
The OK command has one argument, which is the GUID of the server.
See <xref linkend="addresses"/> for more on server GUIDs.
</para>
</sect2>
<sect2 id="auth-command-error">
<title>ERROR Command</title>
<para>
The ERROR command indicates that either server or client did not
know a command, does not accept the given command in the current
context, or did not understand the arguments to the command. This
allows the protocol to be extended; a client or server can send a
command present or permitted only in new protocol versions, and if
an ERROR is received instead of an appropriate response, fall back
to using some other technique.
</para>
<para>
If an ERROR is sent, the server or client that sent the
error must continue as if the command causing the ERROR had never been
received. However, the the server or client receiving the error
should try something other than whatever caused the error;
if only canceling/rejecting the authentication.
</para>
<para>
If the D-Bus protocol changes incompatibly at some future time,
applications implementing the new protocol would probably be able to
check for support of the new protocol by sending a new command and
receiving an ERROR from applications that don't understand it. Thus the
ERROR feature of the auth protocol is an escape hatch that lets us
negotiate extensions or changes to the D-Bus protocol in the future.
</para>
</sect2>
<sect2 id="auth-command-negotiate-unix-fd">
<title>NEGOTIATE_UNIX_FD Command</title>
<para>
The NEGOTIATE_UNIX_FD command indicates that the client
supports Unix file descriptor passing. This command may only
be sent after the connection is authenticated, i.e. after OK
was received by the client. This command may only be sent on
transports that support Unix file descriptor passing.
</para>
<para>
On receiving NEGOTIATE_UNIX_FD the server must respond with
either AGREE_UNIX_FD or ERROR. It shall respond the former if
the transport chosen supports Unix file descriptor passing and
the server supports this feature. It shall respond the latter
if the transport does not support Unix file descriptor
passing, the server does not support this feature, or the
server decides not to enable file descriptor passing due to
security or other reasons.
</para>
</sect2>
<sect2 id="auth-command-agree-unix-fd">
<title>AGREE_UNIX_FD Command</title>
<para>
The AGREE_UNIX_FD command indicates that the server supports
Unix file descriptor passing. This command may only be sent
after the connection is authenticated, and the client sent
NEGOTIATE_UNIX_FD to enable Unix file descriptor passing. This
command may only be sent on transports that support Unix file
descriptor passing.
</para>
<para>
On receiving AGREE_UNIX_FD the client must respond with BEGIN,
followed by its stream of messages, or by disconnecting. The
server must not accept additional commands using this protocol
after the BEGIN command has been received. Further
communication will be a stream of D-Bus messages (optionally
encrypted, as negotiated) rather than this protocol.
</para>
</sect2>
<sect2 id="auth-command-future">
<title>Future Extensions</title>
<para>
Future extensions to the authentication and negotiation
protocol are possible. For that new commands may be
introduced. If a client or server receives an unknown command
it shall respond with ERROR and not consider this fatal. New
commands may be introduced both before, and after
authentication, i.e. both before and after the OK command.
</para>
</sect2>
<sect2 id="auth-examples">
<title>Authentication examples</title>
<para>
<figure>
<title>Example of successful magic cookie authentication</title>
<programlisting>
(MAGIC_COOKIE is a made up mechanism)
C: AUTH MAGIC_COOKIE 3138363935333137393635383634
S: OK 1234deadbeef
C: BEGIN
</programlisting>
</figure>
<figure>
<title>Example of finding out mechanisms then picking one</title>
<programlisting>
C: AUTH
S: REJECTED KERBEROS_V4 SKEY
C: AUTH SKEY 7ab83f32ee
S: DATA 8799cabb2ea93e
C: DATA 8ac876e8f68ee9809bfa876e6f9876g8fa8e76e98f
S: OK 1234deadbeef
C: BEGIN
</programlisting>
</figure>
<figure>
<title>Example of client sends unknown command then falls back to regular auth</title>
<programlisting>
C: FOOBAR
S: ERROR
C: AUTH MAGIC_COOKIE 3736343435313230333039
S: OK 1234deadbeef
C: BEGIN
</programlisting>
</figure>
<figure>
<title>Example of server doesn't support initial auth mechanism</title>
<programlisting>
C: AUTH MAGIC_COOKIE 3736343435313230333039
S: REJECTED KERBEROS_V4 SKEY
C: AUTH SKEY 7ab83f32ee
S: DATA 8799cabb2ea93e
C: DATA 8ac876e8f68ee9809bfa876e6f9876g8fa8e76e98f
S: OK 1234deadbeef
C: BEGIN
</programlisting>
</figure>
<figure>
<title>Example of wrong password or the like followed by successful retry</title>
<programlisting>
C: AUTH MAGIC_COOKIE 3736343435313230333039
S: REJECTED KERBEROS_V4 SKEY
C: AUTH SKEY 7ab83f32ee
S: DATA 8799cabb2ea93e
C: DATA 8ac876e8f68ee9809bfa876e6f9876g8fa8e76e98f
S: REJECTED
C: AUTH SKEY 7ab83f32ee
S: DATA 8799cabb2ea93e
C: DATA 8ac876e8f68ee9809bfa876e6f9876g8fa8e76e98f
S: OK 1234deadbeef
C: BEGIN
</programlisting>
</figure>
<figure>
<title>Example of skey cancelled and restarted</title>
<programlisting>
C: AUTH MAGIC_COOKIE 3736343435313230333039
S: REJECTED KERBEROS_V4 SKEY
C: AUTH SKEY 7ab83f32ee
S: DATA 8799cabb2ea93e
C: CANCEL
S: REJECTED
C: AUTH SKEY 7ab83f32ee
S: DATA 8799cabb2ea93e
C: DATA 8ac876e8f68ee9809bfa876e6f9876g8fa8e76e98f
S: OK 1234deadbeef
C: BEGIN
</programlisting>
</figure>
<figure>
<title>Example of successful magic cookie authentication with successful negotiation of Unix FD passing</title>
<programlisting>
(MAGIC_COOKIE is a made up mechanism)
C: AUTH MAGIC_COOKIE 3138363935333137393635383634
S: OK 1234deadbeef
C: NEGOTIATE_UNIX_FD
S: AGREE_UNIX_FD
C: BEGIN
</programlisting>
</figure>
<figure>
<title>Example of successful magic cookie authentication with unsuccessful negotiation of Unix FD passing</title>
<programlisting>
(MAGIC_COOKIE is a made up mechanism)
C: AUTH MAGIC_COOKIE 3138363935333137393635383634
S: OK 1234deadbeef
C: NEGOTIATE_UNIX_FD
S: ERROR
C: BEGIN
</programlisting>
</figure>
</para>
</sect2>
<sect2 id="auth-states">
<title>Authentication state diagrams</title>
<para>
This section documents the auth protocol in terms of
a state machine for the client and the server. This is
probably the most robust way to implement the protocol.
</para>
<sect3 id="auth-states-client">
<title>Client states</title>
<para>
To more precisely describe the interaction between the
protocol state machine and the authentication mechanisms the
following notation is used: MECH(CHALL) means that the
server challenge CHALL was fed to the mechanism MECH, which
returns one of
<itemizedlist>
<listitem>
<para>
CONTINUE(RESP) means continue the auth conversation
and send RESP as the response to the server;
</para>
</listitem>
<listitem>
<para>
OK(RESP) means that after sending RESP to the server
the client side of the auth conversation is finished
and the server should return "OK";
</para>
</listitem>
<listitem>
<para>
ERROR means that CHALL was invalid and could not be
processed.
</para>
</listitem>
</itemizedlist>
Both RESP and CHALL may be empty.
</para>
<para>
The Client starts by getting an initial response from the
default mechanism and sends AUTH MECH RESP, or AUTH MECH if
the mechanism did not provide an initial response. If the
mechanism returns CONTINUE, the client starts in state
<emphasis>WaitingForData</emphasis>, if the mechanism
returns OK the client starts in state
<emphasis>WaitingForOK</emphasis>.
</para>
<para>
The client should keep track of available mechanisms and
which it mechanisms it has already attempted. This list is
used to decide which AUTH command to send. When the list is
exhausted, the client should give up and close the
connection.
</para>
<formalpara>
<title><emphasis>WaitingForData</emphasis></title>
<para>
<itemizedlist>
<listitem>
<para>
Receive DATA CHALL
<simplelist>
<member>
MECH(CHALL) returns CONTINUE(RESP) &rarr; send
DATA RESP, goto
<emphasis>WaitingForData</emphasis>
</member>
<member>
MECH(CHALL) returns OK(RESP) &rarr; send DATA
RESP, goto <emphasis>WaitingForOK</emphasis>
</member>
<member>
MECH(CHALL) returns ERROR &rarr; send ERROR
[msg], goto <emphasis>WaitingForData</emphasis>
</member>
</simplelist>
</para>
</listitem>
<listitem>
<para>
Receive REJECTED [mechs] &rarr;
send AUTH [next mech], goto
WaitingForData or <emphasis>WaitingForOK</emphasis>
</para>
</listitem>
<listitem>
<para>
Receive ERROR &rarr; send
CANCEL, goto
<emphasis>WaitingForReject</emphasis>
</para>
</listitem>
<listitem>
<para>
Receive OK &rarr; send
BEGIN, terminate auth
conversation, authenticated
</para>
</listitem>
<listitem>
<para>
Receive anything else &rarr; send
ERROR, goto
<emphasis>WaitingForData</emphasis>
</para>
</listitem>
</itemizedlist>
</para>
</formalpara>
<formalpara>
<title><emphasis>WaitingForOK</emphasis></title>
<para>
<itemizedlist>
<listitem>
<para>
Receive OK &rarr; send BEGIN, terminate auth
conversation, <emphasis>authenticated</emphasis>
</para>
</listitem>
<listitem>
<para>
Receive REJECT [mechs] &rarr; send AUTH [next mech],
goto <emphasis>WaitingForData</emphasis> or
<emphasis>WaitingForOK</emphasis>
</para>
</listitem>
<listitem>
<para>
Receive DATA &rarr; send CANCEL, goto
<emphasis>WaitingForReject</emphasis>
</para>
</listitem>
<listitem>
<para>
Receive ERROR &rarr; send CANCEL, goto
<emphasis>WaitingForReject</emphasis>
</para>
</listitem>
<listitem>
<para>
Receive anything else &rarr; send ERROR, goto
<emphasis>WaitingForOK</emphasis>
</para>
</listitem>
</itemizedlist>
</para>
</formalpara>
<formalpara>
<title><emphasis>WaitingForReject</emphasis></title>
<para>
<itemizedlist>
<listitem>
<para>
Receive REJECT [mechs] &rarr; send AUTH [next mech],
goto <emphasis>WaitingForData</emphasis> or
<emphasis>WaitingForOK</emphasis>
</para>
</listitem>
<listitem>
<para>
Receive anything else &rarr; terminate auth
conversation, disconnect
</para>
</listitem>
</itemizedlist>
</para>
</formalpara>
</sect3>
<sect3 id="auth-states-server">
<title>Server states</title>
<para>
For the server MECH(RESP) means that the client response
RESP was fed to the the mechanism MECH, which returns one of
<itemizedlist>
<listitem>
<para>
CONTINUE(CHALL) means continue the auth conversation and
send CHALL as the challenge to the client;
</para>
</listitem>
<listitem>
<para>
OK means that the client has been successfully
authenticated;
</para>
</listitem>
<listitem>
<para>
REJECT means that the client failed to authenticate or
there was an error in RESP.
</para>
</listitem>
</itemizedlist>
The server starts out in state
<emphasis>WaitingForAuth</emphasis>. If the client is
rejected too many times the server must disconnect the
client.
</para>
<formalpara>
<title><emphasis>WaitingForAuth</emphasis></title>
<para>
<itemizedlist>
<listitem>
<para>
Receive AUTH &rarr; send REJECTED [mechs], goto
<emphasis>WaitingForAuth</emphasis>
</para>
</listitem>
<listitem>
<para>
Receive AUTH MECH RESP
<simplelist>
<member>
MECH not valid mechanism &rarr; send REJECTED
[mechs], goto
<emphasis>WaitingForAuth</emphasis>
</member>
<member>
MECH(RESP) returns CONTINUE(CHALL) &rarr; send
DATA CHALL, goto
<emphasis>WaitingForData</emphasis>
</member>
<member>
MECH(RESP) returns OK &rarr; send OK, goto
<emphasis>WaitingForBegin</emphasis>
</member>
<member>
MECH(RESP) returns REJECT &rarr; send REJECTED
[mechs], goto
<emphasis>WaitingForAuth</emphasis>
</member>
</simplelist>
</para>
</listitem>
<listitem>
<para>
Receive BEGIN &rarr; terminate
auth conversation, disconnect
</para>
</listitem>
<listitem>
<para>
Receive ERROR &rarr; send REJECTED [mechs], goto
<emphasis>WaitingForAuth</emphasis>
</para>
</listitem>
<listitem>
<para>
Receive anything else &rarr; send
ERROR, goto
<emphasis>WaitingForAuth</emphasis>
</para>
</listitem>
</itemizedlist>
</para>
</formalpara>
<formalpara>
<title><emphasis>WaitingForData</emphasis></title>
<para>
<itemizedlist>
<listitem>
<para>
Receive DATA RESP
<simplelist>
<member>
MECH(RESP) returns CONTINUE(CHALL) &rarr; send
DATA CHALL, goto
<emphasis>WaitingForData</emphasis>
</member>
<member>
MECH(RESP) returns OK &rarr; send OK, goto
<emphasis>WaitingForBegin</emphasis>
</member>
<member>
MECH(RESP) returns REJECT &rarr; send REJECTED
[mechs], goto
<emphasis>WaitingForAuth</emphasis>
</member>
</simplelist>
</para>
</listitem>
<listitem>
<para>
Receive BEGIN &rarr; terminate auth conversation,
disconnect
</para>
</listitem>
<listitem>
<para>
Receive CANCEL &rarr; send REJECTED [mechs], goto
<emphasis>WaitingForAuth</emphasis>
</para>
</listitem>
<listitem>
<para>
Receive ERROR &rarr; send REJECTED [mechs], goto
<emphasis>WaitingForAuth</emphasis>
</para>
</listitem>
<listitem>
<para>
Receive anything else &rarr; send ERROR, goto
<emphasis>WaitingForData</emphasis>
</para>
</listitem>
</itemizedlist>
</para>
</formalpara>
<formalpara>
<title><emphasis>WaitingForBegin</emphasis></title>
<para>
<itemizedlist>
<listitem>
<para>
Receive BEGIN &rarr; terminate auth conversation,
client authenticated
</para>
</listitem>
<listitem>
<para>
Receive CANCEL &rarr; send REJECTED [mechs], goto
<emphasis>WaitingForAuth</emphasis>
</para>
</listitem>
<listitem>
<para>
Receive ERROR &rarr; send REJECTED [mechs], goto
<emphasis>WaitingForAuth</emphasis>
</para>
</listitem>
<listitem>
<para>
Receive anything else &rarr; send ERROR, goto
<emphasis>WaitingForBegin</emphasis>
</para>
</listitem>
</itemizedlist>
</para>
</formalpara>
</sect3>
</sect2>
<sect2 id="auth-mechanisms">
<title>Authentication mechanisms</title>
<para>
This section describes some new authentication mechanisms.
D-Bus also allows any standard SASL mechanism of course.
</para>
<sect3 id="auth-mechanisms-sha">
<title>DBUS_COOKIE_SHA1</title>
<para>
The DBUS_COOKIE_SHA1 mechanism is designed to establish that a client
has the ability to read a private file owned by the user being
authenticated. If the client can prove that it has access to a secret
cookie stored in this file, then the client is authenticated.
Thus the security of DBUS_COOKIE_SHA1 depends on a secure home
directory.
</para>
<para>
Throughout this description, "hex encoding" must output the digits
from a to f in lower-case; the digits A to F must not be used
in the DBUS_COOKIE_SHA1 mechanism.
</para>
<para>
Authentication proceeds as follows:
<itemizedlist>
<listitem>
<para>
The client sends the username it would like to authenticate
as, hex-encoded.
</para>
</listitem>
<listitem>
<para>
The server sends the name of its "cookie context" (see below); a
space character; the integer ID of the secret cookie the client
must demonstrate knowledge of; a space character; then a
randomly-generated challenge string, all of this hex-encoded into
one, single string.
</para>
</listitem>
<listitem>
<para>
The client locates the cookie and generates its own
randomly-generated challenge string. The client then concatenates
the server's decoded challenge, a ":" character, its own challenge,
another ":" character, and the cookie. It computes the SHA-1 hash
of this composite string as a hex digest. It concatenates the
client's challenge string, a space character, and the SHA-1 hex
digest, hex-encodes the result and sends it back to the server.
</para>
</listitem>
<listitem>
<para>
The server generates the same concatenated string used by the
client and computes its SHA-1 hash. It compares the hash with
the hash received from the client; if the two hashes match, the
client is authenticated.
</para>
</listitem>
</itemizedlist>
</para>
<para>
Each server has a "cookie context," which is a name that identifies a
set of cookies that apply to that server. A sample context might be
"org_freedesktop_session_bus". Context names must be valid ASCII,
nonzero length, and may not contain the characters slash ("/"),
backslash ("\"), space (" "), newline ("\n"), carriage return ("\r"),
tab ("\t"), or period ("."). There is a default context,
"org_freedesktop_general" that's used by servers that do not specify
otherwise.
</para>
<para>
Cookies are stored in a user's home directory, in the directory
<filename>~/.dbus-keyrings/</filename>. This directory must
not be readable or writable by other users. If it is,
clients and servers must ignore it. The directory
contains cookie files named after the cookie context.
</para>
<para>
A cookie file contains one cookie per line. Each line
has three space-separated fields:
<itemizedlist>
<listitem>
<para>
The cookie ID number, which must be a non-negative integer and
may not be used twice in the same file.
</para>
</listitem>
<listitem>
<para>
The cookie's creation time, in UNIX seconds-since-the-epoch
format.
</para>
</listitem>
<listitem>
<para>
The cookie itself, a hex-encoded random block of bytes. The cookie
may be of any length, though obviously security increases
as the length increases.
</para>
</listitem>
</itemizedlist>
</para>
<para>
Only server processes modify the cookie file.
They must do so with this procedure:
<itemizedlist>
<listitem>
<para>
Create a lockfile name by appending ".lock" to the name of the
cookie file. The server should attempt to create this file
using <literal>O_CREAT | O_EXCL</literal>. If file creation
fails, the lock fails. Servers should retry for a reasonable
period of time, then they may choose to delete an existing lock
to keep users from having to manually delete a stale
lock. <footnote><para>Lockfiles are used instead of real file
locking <literal>fcntl()</literal> because real locking
implementations are still flaky on network
filesystems.</para></footnote>
</para>
</listitem>
<listitem>
<para>
Once the lockfile has been created, the server loads the cookie
file. It should then delete any cookies that are old (the
timeout can be fairly short), or more than a reasonable
time in the future (so that cookies never accidentally
become permanent, if the clock was set far into the future
at some point). If no recent keys remain, the
server may generate a new key.
</para>
</listitem>
<listitem>
<para>
The pruned and possibly added-to cookie file
must be resaved atomically (using a temporary
file which is rename()'d).
</para>
</listitem>
<listitem>
<para>
The lock must be dropped by deleting the lockfile.
</para>
</listitem>
</itemizedlist>
</para>
<para>
Clients need not lock the file in order to load it,
because servers are required to save the file atomically.
</para>
</sect3>
</sect2>
</sect1>
<sect1 id="addresses">
<title>Server Addresses</title>
<para>
Server addresses consist of a transport name followed by a colon, and
then an optional, comma-separated list of keys and values in the form key=value.
Each value is escaped.
</para>
<para>
For example:
<programlisting>unix:path=/tmp/dbus-test</programlisting>
Which is the address to a unix socket with the path /tmp/dbus-test.
</para>
<para>
Value escaping is similar to URI escaping but simpler.
<itemizedlist>
<listitem>
<para>
The set of optionally-escaped bytes is:
<literal>[0-9A-Za-z_-/.\]</literal>. To escape, each
<emphasis>byte</emphasis> (note, not character) which is not in the
set of optionally-escaped bytes must be replaced with an ASCII
percent (<literal>%</literal>) and the value of the byte in hex.
The hex value must always be two digits, even if the first digit is
zero. The optionally-escaped bytes may be escaped if desired.
</para>
</listitem>
<listitem>
<para>
To unescape, append each byte in the value; if a byte is an ASCII
percent (<literal>%</literal>) character then append the following
hex value instead. It is an error if a <literal>%</literal> byte
does not have two hex digits following. It is an error if a
non-optionally-escaped byte is seen unescaped.
</para>
</listitem>
</itemizedlist>
The set of optionally-escaped bytes is intended to preserve address
readability and convenience.
</para>
<para>
A server may specify a key-value pair with the key <literal>guid</literal>
and the value a hex-encoded 16-byte sequence. <xref linkend="uuids"/>
describes the format of the <literal>guid</literal> field. If present,
this UUID may be used to distinguish one server address from another. A
server should use a different UUID for each address it listens on. For
example, if a message bus daemon offers both UNIX domain socket and TCP
connections, but treats clients the same regardless of how they connect,
those two connections are equivalent post-connection but should have
distinct UUIDs to distinguish the kinds of connection.
</para>
<para>
The intent of the address UUID feature is to allow a client to avoid
opening multiple identical connections to the same server, by allowing the
client to check whether an address corresponds to an already-existing
connection. Comparing two addresses is insufficient, because addresses
can be recycled by distinct servers, and equivalent addresses may look
different if simply compared as strings (for example, the host in a TCP
address can be given as an IP address or as a hostname).
</para>
<para>
Note that the address key is <literal>guid</literal> even though the
rest of the API and documentation says "UUID," for historical reasons.
</para>
<para>
[FIXME clarify if attempting to connect to each is a requirement
or just a suggestion]
When connecting to a server, multiple server addresses can be
separated by a semi-colon. The library will then try to connect
to the first address and if that fails, it'll try to connect to
the next one specified, and so forth. For example
<programlisting>unix:path=/tmp/dbus-test;unix:path=/tmp/dbus-test2</programlisting>
</para>
</sect1>
<sect1 id="transports">
<title>Transports</title>
<para>
[FIXME we need to specify in detail each transport and its possible arguments]
Current transports include: unix domain sockets (including
abstract namespace on linux), launchd, systemd, TCP/IP, an executed subprocess and a debug/testing transport
using in-process pipes. Future possible transports include one that
tunnels over X11 protocol.
</para>
<sect2 id="transports-unix-domain-sockets">
<title>Unix Domain Sockets</title>
<para>
Unix domain sockets can be either paths in the file system or on Linux
kernels, they can be abstract which are similar to paths but
do not show up in the file system.
</para>
<para>
When a socket is opened by the D-Bus library it truncates the path
name right before the first trailing Nul byte. This is true for both
normal paths and abstract paths. Note that this is a departure from
previous versions of D-Bus that would create sockets with a fixed
length path name. Names which were shorter than the fixed length
would be padded by Nul bytes.
</para>
<para>
Unix domain sockets are not available on Windows.
</para>
<sect3 id="transports-unix-domain-sockets-addresses">
<title>Server Address Format</title>
<para>
Unix domain socket addresses are identified by the "unix:" prefix
and support the following key/value pairs:
</para>
<informaltable>
<tgroup cols="3">
<thead>
<row>
<entry>Name</entry>
<entry>Values</entry>
<entry>Description</entry>
</row>
</thead>
<tbody>
<row>
<entry>path</entry>
<entry>(path)</entry>
<entry>path of the unix domain socket. If set, the "tmpdir" and "abstract" key must not be set.</entry>
</row>
<row>
<entry>tmpdir</entry>
<entry>(path)</entry>
<entry>temporary directory in which a socket file with a random file name starting with 'dbus-' will be created by the server. This key can only be used in server addresses, not in client addresses. If set, the "path" and "abstract" key must not be set.</entry>
</row>
<row>
<entry>abstract</entry>
<entry>(string)</entry>
<entry>unique string (path) in the abstract namespace. If set, the "path" or "tempdir" key must not be set.</entry>
</row>
</tbody>
</tgroup>
</informaltable>
</sect3>
</sect2>
<sect2 id="transports-launchd">
<title>launchd</title>
<para>
launchd is an open-source server management system that replaces init, inetd
and cron on Apple Mac OS X versions 10.4 and above. It provides a common session
bus address for each user and deprecates the X11-enabled D-Bus launcher on OSX.
</para>
<para>
launchd allocates a socket and provides it with the unix path through the
DBUS_LAUNCHD_SESSION_BUS_SOCKET variable in launchd's environment. Every process
spawned by launchd (or dbus-daemon, if it was started by launchd) can access
it through its environment.
Other processes can query for the launchd socket by executing:
$ launchctl getenv DBUS_LAUNCHD_SESSION_BUS_SOCKET
This is normally done by the D-Bus client library so doesn't have to be done
manually.
</para>
<para>
launchd is not available on Microsoft Windows.
</para>
<sect3 id="transports-launchd-addresses">
<title>Server Address Format</title>
<para>
launchd addresses are identified by the "launchd:" prefix
and support the following key/value pairs:
</para>
<informaltable>
<tgroup cols="3">
<thead>
<row>
<entry>Name</entry>
<entry>Values</entry>
<entry>Description</entry>
</row>
</thead>
<tbody>
<row>
<entry>env</entry>
<entry>(environment variable)</entry>
<entry>path of the unix domain socket for the launchd created dbus-daemon.</entry>
</row>
</tbody>
</tgroup>
</informaltable>
</sect3>
</sect2>
<sect2 id="transports-systemd">
<title>systemd</title>
<para>
systemd is an open-source server management system that
replaces init and inetd on newer Linux systems. It supports
socket activation. The D-Bus systemd transport is used to acquire
socket activation file descriptors from systemd and use them
as D-Bus transport when the current process is spawned by
socket activation from it.
</para>
<para>
The systemd transport accepts only one or more Unix domain or
TCP streams sockets passed in via socket activation.
</para>
<para>
The systemd transport is not available on non-Linux operating systems.
</para>
<para>
The systemd transport defines no parameter keys.
</para>
</sect2>
<sect2 id="transports-tcp-sockets">
<title>TCP Sockets</title>
<para>
The tcp transport provides TCP/IP based connections between clients
located on the same or different hosts.
</para>
<para>
Using tcp transport without any additional secure authentification mechanismus
over a network is unsecure.
</para>
<para>
Windows notes: Because of the tcp stack on Windows does not provide sending
credentials over a tcp connection, the EXTERNAL authentification
mechanismus does not work.
</para>
<sect3 id="transports-tcp-sockets-addresses">
<title>Server Address Format</title>
<para>
TCP/IP socket addresses are identified by the "tcp:" prefix
and support the following key/value pairs:
</para>
<informaltable>
<tgroup cols="3">
<thead>
<row>
<entry>Name</entry>
<entry>Values</entry>
<entry>Description</entry>
</row>
</thead>
<tbody>
<row>
<entry>host</entry>
<entry>(string)</entry>
<entry>dns name or ip address</entry>
</row>
<row>
<entry>port</entry>
<entry>(number)</entry>
<entry>The tcp port the server will open. A zero value let the server
choose a free port provided from the underlaying operating system.
libdbus is able to retrieve the real used port from the server.
</entry>
</row>
<row>
<entry>family</entry>
<entry>(string)</entry>
<entry>If set, provide the type of socket family either "ipv4" or "ipv6". If unset, the family is unspecified.</entry>
</row>
</tbody>
</tgroup>
</informaltable>
</sect3>
</sect2>
<sect2 id="transports-nonce-tcp-sockets">
<title>Nonce-secured TCP Sockets</title>
<para>
The nonce-tcp transport provides a secured TCP transport, using a
simple authentication mechanism to ensure that only clients with read
access to a certain location in the filesystem can connect to the server.
The server writes a secret, the nonce, to a file and an incoming client
connection is only accepted if the client sends the nonce right after
the connect. The nonce mechanism requires no setup and is orthogonal to
the higher-level authentication mechanisms described in the
Authentication section.
</para>
<para>
On start, the server generates a random 16 byte nonce and writes it
to a file in the user's temporary directory. The nonce file location
is published as part of the server's D-Bus address using the
"noncefile" key-value pair.
After an accept, the server reads 16 bytes from the socket. If the
read bytes do not match the nonce stored in the nonce file, the
server MUST immediately drop the connection.
If the nonce match the received byte sequence, the client is accepted
and the transport behaves like an unsecured tcp transport.
</para>
<para>
After a successful connect to the server socket, the client MUST read
the nonce from the file published by the server via the noncefile=
key-value pair and send it over the socket. After that, the
transport behaves like an unsecured tcp transport.
</para>
<sect3 id="transports-nonce-tcp-sockets-addresses">
<title>Server Address Format</title>
<para>
Nonce TCP/IP socket addresses uses the "nonce-tcp:" prefix
and support the following key/value pairs:
</para>
<informaltable>
<tgroup cols="3">
<thead>
<row>
<entry>Name</entry>
<entry>Values</entry>
<entry>Description</entry>
</row>
</thead>
<tbody>
<row>
<entry>host</entry>
<entry>(string)</entry>
<entry>dns name or ip address</entry>
</row>
<row>
<entry>port</entry>
<entry>(number)</entry>
<entry>The tcp port the server will open. A zero value let the server
choose a free port provided from the underlaying operating system.
libdbus is able to retrieve the real used port from the server.
</entry>
</row>
<row>
<entry>family</entry>
<entry>(string)</entry>
<entry>If set, provide the type of socket family either "ipv4" or "ipv6". If unset, the family is unspecified.</entry>
</row>
<row>
<entry>noncefile</entry>
<entry>(path)</entry>
<entry>file location containing the secret</entry>
</row>
</tbody>
</tgroup>
</informaltable>
</sect3>
</sect2>
<sect2 id="transports-exec">
<title>Executed Subprocesses on Unix</title>
<para>
This transport forks off a process and connects its standard
input and standard output with an anonymous Unix domain
socket. This socket is then used for communication by the
transport. This transport may be used to use out-of-process
forwarder programs as basis for the D-Bus protocol.
</para>
<para>
The forked process will inherit the standard error output and
process group from the parent process.
</para>
<para>
Executed subprocesses are not available on Windows.
</para>
<sect3 id="transports-exec-addresses">
<title>Server Address Format</title>
<para>
Executed subprocess addresses are identified by the "unixexec:" prefix
and support the following key/value pairs:
</para>
<informaltable>
<tgroup cols="3">
<thead>
<row>
<entry>Name</entry>
<entry>Values</entry>
<entry>Description</entry>
</row>
</thead>
<tbody>
<row>
<entry>path</entry>
<entry>(path)</entry>
<entry>Path of the binary to execute, either an absolute
path or a binary name that is searched for in the default
search path of the OS. This corresponds to the first
argument of execlp(). This key is mandatory.</entry>
</row>
<row>
<entry>argv0</entry>
<entry>(string)</entry>
<entry>The program name to use when executing the
binary. If omitted the same value as specified for path=
will be used. This corresponds to the second argument of
execlp().</entry>
</row>
<row>
<entry>argv1, argv2, ...</entry>
<entry>(string)</entry>
<entry>Arguments to pass to the binary. This corresponds
to the third and later arguments of execlp(). If a
specific argvX is not specified no further argvY for Y > X
are taken into account.</entry>
</row>
</tbody>
</tgroup>
</informaltable>
</sect3>
</sect2>
</sect1>
<sect1 id="meta-transports">
<title>Meta Transports</title>
<para>
Meta transports are a kind of transport with special enhancements or
behavior. Currently available meta transports include: autolaunch
</para>
<sect2 id="meta-transports-autolaunch">
<title>Autolaunch</title>
<para>The autolaunch transport provides a way for dbus clients to autodetect
a running dbus session bus and to autolaunch a session bus if not present.
</para>
<sect3 id="meta-transports-autolaunch-addresses">
<title>Server Address Format</title>
<para>
Autolaunch addresses uses the "autolaunch:" prefix and support the
following key/value pairs:
</para>
<informaltable>
<tgroup cols="3">
<thead>
<row>
<entry>Name</entry>
<entry>Values</entry>
<entry>Description</entry>
</row>
</thead>
<tbody>
<row>
<entry>scope</entry>
<entry>(string)</entry>
<entry>scope of autolaunch (Windows only)
<itemizedlist>
<listitem>
<para>
"*install-path" - limit session bus to dbus installation path.
The dbus installation path is determined from the location of
the shared dbus library. If the library is located in a 'bin'
subdirectory the installation root is the directory above,
otherwise the directory where the library lives is taken as
installation root.
<programlisting>
&lt;install-root&gt;/bin/[lib]dbus-1.dll
&lt;install-root&gt;/[lib]dbus-1.dll
</programlisting>
</para>
</listitem>
<listitem>
<para>
"*user" - limit session bus to the recent user.
</para>
</listitem>
<listitem>
<para>
other values - specify dedicated session bus like "release",
"debug" or other
</para>
</listitem>
</itemizedlist>
</entry>
</row>
</tbody>
</tgroup>
</informaltable>
</sect3>
<sect3 id="meta-transports-autolaunch-windows-implementation">
<title>Windows implementation</title>
<para>
On start, the server opens a platform specific transport, creates a mutex
and a shared memory section containing the related session bus address.
This mutex will be inspected by the dbus client library to detect a
running dbus session bus. The access to the mutex and the shared memory
section are protected by global locks.
</para>
<para>
In the recent implementation the autolaunch transport uses a tcp transport
on localhost with a port choosen from the operating system. This detail may
change in the future.
</para>
<para>
Disclaimer: The recent implementation is in an early state and may not
work in all cirumstances and/or may have security issues. Because of this
the implementation is not documentated yet.
</para>
</sect3>
</sect2>
</sect1>
<sect1 id="uuids">
<title>UUIDs</title>
<para>
A working D-Bus implementation uses universally-unique IDs in two places.
First, each server address has a UUID identifying the address,
as described in <xref linkend="addresses"/>. Second, each operating
system kernel instance running a D-Bus client or server has a UUID
identifying that kernel, retrieved by invoking the method
org.freedesktop.DBus.Peer.GetMachineId() (see <xref
linkend="standard-interfaces-peer"/>).
</para>
<para>
The term "UUID" in this document is intended literally, i.e. an
identifier that is universally unique. It is not intended to refer to
RFC4122, and in fact the D-Bus UUID is not compatible with that RFC.
</para>
<para>
The UUID must contain 128 bits of data and be hex-encoded. The
hex-encoded string may not contain hyphens or other non-hex-digit
characters, and it must be exactly 32 characters long. To generate a
UUID, the current reference implementation concatenates 96 bits of random
data followed by the 32-bit time in seconds since the UNIX epoch (in big
endian byte order).
</para>
<para>
It would also be acceptable and probably better to simply generate 128
bits of random data, as long as the random number generator is of high
quality. The timestamp could conceivably help if the random bits are not
very random. With a quality random number generator, collisions are
extremely unlikely even with only 96 bits, so it's somewhat academic.
</para>
<para>
Implementations should, however, stick to random data for the first 96 bits
of the UUID.
</para>
</sect1>
<sect1 id="standard-interfaces">
<title>Standard Interfaces</title>
<para>
See <xref linkend="message-protocol-types-notation"/> for details on
the notation used in this section. There are some standard interfaces
that may be useful across various D-Bus applications.
</para>
<sect2 id="standard-interfaces-peer">
<title><literal>org.freedesktop.DBus.Peer</literal></title>
<para>
The <literal>org.freedesktop.DBus.Peer</literal> interface
has two methods:
<programlisting>
org.freedesktop.DBus.Peer.Ping ()
org.freedesktop.DBus.Peer.GetMachineId (out STRING machine_uuid)
</programlisting>
</para>
<para>
On receipt of the <literal>METHOD_CALL</literal> message
<literal>org.freedesktop.DBus.Peer.Ping</literal>, an application should do
nothing other than reply with a <literal>METHOD_RETURN</literal> as
usual. It does not matter which object path a ping is sent to. The
reference implementation handles this method automatically.
</para>
<para>
On receipt of the <literal>METHOD_CALL</literal> message
<literal>org.freedesktop.DBus.Peer.GetMachineId</literal>, an application should
reply with a <literal>METHOD_RETURN</literal> containing a hex-encoded
UUID representing the identity of the machine the process is running on.
This UUID must be the same for all processes on a single system at least
until that system next reboots. It should be the same across reboots
if possible, but this is not always possible to implement and is not
guaranteed.
It does not matter which object path a GetMachineId is sent to. The
reference implementation handles this method automatically.
</para>
<para>
The UUID is intended to be per-instance-of-the-operating-system, so may represent
a virtual machine running on a hypervisor, rather than a physical machine.
Basically if two processes see the same UUID, they should also see the same
shared memory, UNIX domain sockets, process IDs, and other features that require
a running OS kernel in common between the processes.
</para>
<para>
The UUID is often used where other programs might use a hostname. Hostnames
can change without rebooting, however, or just be "localhost" - so the UUID
is more robust.
</para>
<para>
<xref linkend="uuids"/> explains the format of the UUID.
</para>
</sect2>
<sect2 id="standard-interfaces-introspectable">
<title><literal>org.freedesktop.DBus.Introspectable</literal></title>
<para>
This interface has one method:
<programlisting>
org.freedesktop.DBus.Introspectable.Introspect (out STRING xml_data)
</programlisting>
</para>
<para>
Objects instances may implement
<literal>Introspect</literal> which returns an XML description of
the object, including its interfaces (with signals and methods), objects
below it in the object path tree, and its properties.
</para>
<para>
<xref linkend="introspection-format"/> describes the format of this XML string.
</para>
</sect2>
<sect2 id="standard-interfaces-properties">