wayland/doc/publican/sources/Protocol.xml

<?xml version='1.0' encoding='utf-8' ?>
<!DOCTYPE chapter PUBLIC "-//OASIS//DTD DocBook XML V4.5//EN" "http://www.oasis-open.org/docbook/xml/4.5/docbookx.dtd" [
<!ENTITY % BOOK_ENTITIES SYSTEM "Wayland.ent">
%BOOK_ENTITIES;
]>
<chapter id="chap-Protocol">
  <title>Wayland Protocol and Model of Operation</title>
  <section id="sect-Protocol-Basic-Principles">
    <title>Basic Principles</title>
    <para>
      The Wayland protocol is an asynchronous object oriented protocol.  All
      requests are method invocations on some object.  The requests include
      an object ID that uniquely identifies an object on the server.  Each
      object implements an interface and the requests include an opcode that
      identifies which method in the interface to invoke.
    </para>
    <para>
      The server sends back events to the client, each event is emitted from
      an object.  Events can be error conditions.  The event includes the
      object ID and the event opcode, from which the client can determine
      the type of event.  Events are generated both in response to requests
      (in which case the request and the event constitutes a round trip) or
      spontaneously when the server state changes.
    </para>
    <para>
      <itemizedlist>
	<listitem>
	  <para>
	    State is broadcast on connect, events are sent
	    out when state changes. Clients must listen for
	    these changes and cache the state.
	    There is no need (or mechanism) to query server state.
	  </para>
	</listitem>
	<listitem>
	  <para>
	    The server will broadcast the presence of a number of global objects,
	    which in turn will broadcast their current state.
	  </para>
	</listitem>
      </itemizedlist>
    </para>
  </section>
  <section id="sect-Protocol-Code-Generation">
    <title>Code Generation</title>
    <para>
      The interfaces, requests and events are defined in
      <filename>protocol/wayland.xml</filename>.
      This xml is used to generate the function prototypes that can be used by
      clients and compositors.
    </para>
    <para>
      The protocol entry points are generated as inline functions which just
      wrap the <function>wl_proxy_*</function> functions.  The inline functions aren't
      part of the library ABI and language bindings should generate their
      own stubs for the protocol entry points from the xml.
    </para>
  </section>
  <section id="sect-Protocol-Wire-Format">
    <title>Wire Format</title>
    <para>
      The protocol is sent over a UNIX domain stream socket, where the endpoint
      usually is named <systemitem class="service">wayland-0</systemitem>
      (although it can be changed via <emphasis>WAYLAND_DISPLAY</emphasis>
      in the environment).  The protocol is message-based.  A
      message sent by a client to the server is called request.  A message
      from the server to a client is called event.  Every message is
      structured as 32-bit words, values are represented in the host's
      byte-order.
    </para>
    <para>
      The message header has 2 words in it:
      <itemizedlist>
	<listitem>
	  <para>
	    The first word is the sender's object ID (32-bit).
	  </para>
	</listitem>
	<listitem>
	  <para>
	    The second has 2 parts of 16-bit.  The upper 16-bits are the message
	    size in bytes, starting at the header (i.e. it has a minimum value of 8).The lower is the request/event opcode.
	  </para>
	</listitem>
      </itemizedlist>
      The payload describes the request/event arguments.  Every argument is always
      aligned to 32-bits. Where padding is required, the value of padding bytes is
      undefined. There is no prefix that describes the type, but it is
      inferred implicitly from the xml specification.
    </para>
    <para>

      The representation of argument types are as follows:
      <variablelist>
	<varlistentry>
	  <term>int</term>
	  <term>uint</term>
	  <listitem>
	    <para>
	      The value is the 32-bit value of the signed/unsigned
	      int.
	    </para>
	  </listitem>
	</varlistentry>
	<varlistentry>
	  <term>fixed</term>
	  <listitem>
	    <para>
	      Signed 24.8 decimal numbers. It is a signed decimal type which
	      offers a sign bit, 23 bits of integer precision and 8 bits of
	      decimal precision. This is exposed as an opaque struct with
	      conversion helpers to and from double and int on the C API side.
	    </para>
	  </listitem>
	</varlistentry>
	<varlistentry>
	  <term>string</term>
	  <listitem>
	    <para>
	      Starts with an unsigned 32-bit length, followed by the
	      string contents, including terminating null byte, then padding
	      to a 32-bit boundary.
	    </para>
	  </listitem>
	</varlistentry>
	<varlistentry>
	  <term>object</term>
	  <listitem>
	    <para>
	      32-bit object ID.
	    </para>
	  </listitem>
	</varlistentry>
	<varlistentry>
	  <term>new_id</term>
	  <listitem>
	    <para>
	      The 32-bit object ID.  On requests, the client
	      decides the ID.  The only events with <type>new_id</type> are
	      advertisements of globals, and the server will use IDs below
	      0x10000.
	    </para>
	  </listitem>
	</varlistentry>
	<varlistentry>
	  <term>array</term>
	  <listitem>
	    <para>
	      Starts with 32-bit array size in bytes, followed by the array
	      contents verbatim, and finally padding to a 32-bit boundary.
	    </para>
	  </listitem>
	</varlistentry>
	<varlistentry>
	  <term>fd</term>
	  <listitem>
	    <para>
	      The file descriptor is not stored in the message buffer, but in
	      the ancillary data of the UNIX domain socket message (msg_control).
	    </para>
	  </listitem>
	</varlistentry>
      </variablelist>
    </para>
  </section>
  <xi:include href="ProtocolInterfaces.xml" xmlns:xi="http://www.w3.org/2001/XInclude"/>
  <section id="sect-Protocol-Versioning">
    <title>Versioning</title>
    <para>
      Every interface is versioned and every protocol object implements a
      particular version of its interface.  For global objects, the maximum
      version supported by the server is advertised with the global and the
      actual verion of the created protocol object is determined by the
      version argument passed to wl_registry.bind().  For objects that are
      not globals, their version is inferred from the object that created
      them.
    </para>
    <para>
      In order to keep things sane, this has a few implications for
      interface versions:
      <itemizedlist>
	<listitem>
	  <para>
	    The object creation hierarchy must be a tree.  Otherwise,
	    infering object versions from the parent object becomes a much
	    more difficult to properly track.
	  </para>
	</listitem>
	<listitem>
	  <para>
	    When the version of an interface increases, so does the version
	    of its parent (recursively until you get to a global interface)
	  </para>
	</listitem>
	<listitem>
	  <para>
	    A global interface's version number acts like a counter for all
	    of its child interfaces.  Whenever a child interface gets
	    modified, the global parent's interface version number also
	    increases (see above).  The child interface then takes on the
	    same version number as the new version of its parent global
	    interface.
	  </para>
	</listitem>
      </itemizedlist>
    </para>
    <para>
      To illustrate the above, consider the wl_compositor interface.  It
      has two children, wl_surface and wl_region.  As of wayland version
      1.2, wl_surface and wl_compositor are both at version 3.  If
      something is added to the wl_region interface, both wl_region and
      wl_compositor will get bumpped to version 4.  If, afterwards,
      wl_surface is changed, both wl_compositor and wl_surface will be at
      version 5.  In this way the global interface version is used as a
      sort of "counter" for all of its child interfaces.  This makes it
      very simple to know the version of the child given the version of its
      parent.  The child is at the highest possible interface version that
      is less than or equal to its parent's version.
    </para>
    <para>
      It is worth noting a particular exception to the above versioning
      scheme.  The wl_display (and, by extension, wl_registry) interface
      cannot change because it is the core protocol object and its version
      is never advertised nor is there a mechanism to request a different
      version.
    </para>
  </section>
  <section id="sect-Protocol-Connect-Time">
    <title>Connect Time</title>
    <para>
      There is no fixed connection setup information, the server emits
      multiple events at connect time, to indicate the presence and
      properties of global objects: outputs, compositor, input devices.
    </para>
  </section>
  <section id="sect-Protocol-Security-and-Authentication">
    <title>Security and Authentication</title>
    <para>
      <itemizedlist>
	<listitem>
	  <para>
	    mostly about access to underlying buffers, need new drm auth
	    mechanism (the grant-to ioctl idea), need to check the cmd stream?
	  </para>
	</listitem>
	<listitem>
	  <para>
	    getting the server socket depends on the compositor type, could
	    be a system wide name, through fd passing on the session dbus.
	    or the client is forked by the compositor and the fd is
	    already opened.
	  </para>
	</listitem>
      </itemizedlist>
    </para>
  </section>
  <section id="sect-Protocol-Creating-Objects">
    <title>Creating Objects</title>
    <para>
      Each object has a unique ID.  The IDs are allocated by the entity
      creating the object (either client or server).  IDs allocated by the
      client are in the range [1, 0xfeffffff] while IDs allocated by the
      server are in the range [0xff000000, 0xffffffff].  The 0 ID is
      reserved to represent a null or non-existant object.

      For efficiency purposes, the IDs are densely packed in the sense that
      the ID N will not be used until N-1 has been used.  Any ID allocation
      algorithm that does not maintain this property is incompatible with
      the implementation in libwayland.
    </para>
  </section>
  <section id="sect-Protocol-Compositor">
    <title>Compositor</title>
    <para>
      The compositor is a global object, advertised at connect time.
    </para>
    <para>
      See <xref linkend="protocol-spec-wl_compositor"/> for the
      protocol description.
    </para>
  </section>
  <section id="sect-Protocol-Surface">
    <title>Surfaces</title>
    <para>
      Surfaces are created by the client.
      Clients don't know the global position of their surfaces, and
      cannot access other clients surfaces.
    </para>
    <para>
      See <xref linkend="protocol-spec-wl_surface"/> for the protocol
      description.
    </para>
  </section>
  <section id="sect-Protocol-Input">
    <title>Input</title>
    <para>
      A seat represents a group of input devices including mice,
      keyboards and touchscreens. It has a keyboard and pointer
      focus. Seats are global objects. Pointer events are delivered
      in surface local coordinates.
    </para>
    <para>
      The compositor maintains an implicit grab when a button is
      pressed, to ensure that the corresponding button release
      event gets delivered to the same surface. But there is no way
      for clients to take an explicit grab. Instead, surfaces can
      be mapped as 'popup', which combines transient window semantics
      with a pointer grab.
    </para>
    <para>
      To avoid race conditions, input events that are likely to
      trigger further requests (such as button presses, key events,
      pointer motions) carry serial numbers, and requests such as
      wl_surface.set_popup require that the serial number of the
      triggering event is specified. The server maintains a
      monotonically increasing counter for these serial numbers.
    </para>
    <para>
      Input events also carry timestamps with millisecond granularity.
      Their base is undefined, so they can't be compared against
      system time (as obtained with clock_gettime or gettimeofday).
      They can be compared with each other though, and for instance
      be used to identify sequences of button presses as double
      or triple clicks.
    </para>
    <para>
      See <xref linkend="protocol-spec-wl_seat"/> for the
      protocol description.
    </para>
    <para>
      Talk about:

      <itemizedlist>
	<listitem>
	  <para>
	    keyboard map, change events
	  </para>
	</listitem>
	<listitem>
	  <para>
	    xkb on Wayland
	  </para>
	</listitem>
	<listitem>
	  <para>
	    multi pointer Wayland
	  </para>
	</listitem>
      </itemizedlist>
    </para>
    <para>
      A surface can change the pointer image when the surface is the pointer
      focus of the input device.  Wayland doesn't automatically change the
      pointer image when a pointer enters a surface, but expects the
      application to set the cursor it wants in response to the pointer
      focus and motion events.  The rationale is that a client has to manage
      changing pointer images for UI elements within the surface in response
      to motion events anyway, so we'll make that the only mechanism for
      setting or changing the pointer image.  If the server receives a request
      to set the pointer image after the surface loses pointer focus, the
      request is ignored.  To the client this will look like it successfully
      set the pointer image.
    </para>
    <para>
      The compositor will revert the pointer image back to a default image
      when no surface has the pointer focus for that device.  Clients can
      revert the pointer image back to the default image by setting a NULL
      image.
    </para>
    <para>
      What if the pointer moves from one window which has set a special
      pointer image to a surface that doesn't set an image in response to
      the motion event?  The new surface will be stuck with the special
      pointer image.  We can't just revert the pointer image on leaving a
      surface, since if we immediately enter a surface that sets a different
      image, the image will flicker.  Broken app, I suppose.
    </para>
  </section>
  <section id="sect-Protocol-Output">
    <title>Output</title>
    <para>
      An output is a global object, advertised at connect time or as it
      comes and goes.
    </para>
    <para>
      See <xref linkend="protocol-spec-wl_output"/> for the protocol
      description.
    </para>
    <para>
    </para>
    <itemizedlist>
      <listitem>
	<para>
	  laid out in a big (compositor) coordinate system
	</para>
      </listitem>
      <listitem>
	<para>
	  basically xrandr over Wayland
	</para>
      </listitem>
      <listitem>
	<para>
	  geometry needs position in compositor coordinate system
	</para>
      </listitem>
      <listitem>
	<para>
	  events to advertise available modes, requests to move and change
	  modes
	</para>
      </listitem>
    </itemizedlist>
  </section>
  <section id="sect-Protocol-data-sharing">
    <title>Data sharing between clients</title>
    <para>
      The Wayland protocol provides clients a mechanism for sharing
      data that allows the implementation of copy-paste and
      drag-and-drop. The client providing the data creates a
      <function>wl_data_source</function> object and the clients
      obtaining the data will see it as <function>wl_data_offer</function>
      object. This interface allows the clients to agree on a mutually
      supported mime type and transfer the data via a file descriptor
      that is passed through the protocol.
    </para>
    <para>
      The next section explains the negotiation between data source and
      data offer objects. <xref linkend="sect-Protocol-data-sharing-devices"/>
      explains how these objects are created and passed to different
      clients using the <function>wl_data_device</function> interface
      that implements copy-paste and drag-and-drop support.
    </para>
    <para>
      See <xref linkend="protocol-spec-wl_data_offer"/>,
      <xref linkend="protocol-spec-wl_data_source"/>,
      <xref linkend="protocol-spec-wl_data_device"/> and
      <xref linkend="protocol-spec-wl_data_device_manager"/> for
      protocol descriptions.
    </para>
    <para>
      MIME is defined in RFC's 2045-2049. A
      <ulink url="ftp://ftp.isi.edu/in-notes/iana/assignments/media-types/">
      registry of MIME types</ulink> is maintained by the Internet Assigned
      Numbers Authority (IANA).
    </para>
    <section>
      <title>Data negotiation</title>
      <para>
	A client providing data to other clients will create a <function>wl_data_source</function>
	object and advertise the mime types for the formats it supports for
	that data through the <function>wl_data_source.offer</function>
	request. On the receiving end, the data offer object will generate one
	<function>wl_data_offer.offer</function> event for each supported mime
	type.
      </para>
      <para>
	The actual data transfer happens when the receiving client sends a
	<function>wl_data_offer.receive</function> request. This request takes
	a mime type and a file descriptor as arguments. This request will generate a
	<function>wl_data_source.send</function> event on the sending client
	with the same arguments, and the latter client is expected to write its
	data to the given file descriptor using the chosen mime type.
      </para>
    </section>
    <section id="sect-Protocol-data-sharing-devices">
      <title>Data devices</title>
      <para>
	Data devices glue data sources and offers together. A data device is
	associated with a <function>wl_seat</function> and is obtained by the clients using the
	<function>wl_data_device_manager</function> factory object, which is also responsible for
	creating data sources.
      </para>
      <para>
	Clients are informed of new data offers through the
	<function>wl_data_device.data_offer</function> event. After this
	event is generated the data offer will advertise the available mime
	types. New data offers are introduced prior to their use for
	copy-paste or drag-and-drop.
      </para>
      <section>
	<title>Selection</title>
	<para>
	  Each data device has a selection data source. Clients create a data
	  source object using the device manager and may set it as the
	  current selection for a given data device. Whenever the current
	  selection changes, the client with keyboard focus receives a
	  <function>wl_data_device.selection</function> event. This event is
	  also generated on a client immediately before it receives keyboard
	  focus.
	</para>
	<para>
	  The data offer is introduced with
	  <function>wl_data_device.data_offer</function> event before the
	  selection event.
	</para>
      </section>
      <section>
	<title>Drag and Drop</title>
	<para>
	  A drag-and-drop operation is started using the
	  <function>wl_data_device.start_drag</function> request. This
	  requests causes a pointer grab that will generate enter, motion and
	  leave events on the data device. A data source is supplied as
	  argument to start_drag, and data offers associated with it are
	  supplied to clients surfaces under the pointer in the
	  <function>wl_data_device.enter</function> event. The data offer
	  is introduced to the client prior to the enter event with the
	  <function>wl_data_device.data_offer</function> event.
	</para>
	<para>
	  Clients are expected to provide feedback to the data sending client
	  by calling the <function>wl_data_offer.accept</function> request with
	  a mime type it accepts. If none of the advertised mime types is
	  supported by the receiving client, it should supply NULL to the
	  accept request. The accept request causes the sending client to
	  receive a <function>wl_data_source.target</function> event with the
	  chosen mime type.
	</para>
	<para>
	  When the drag ends, the receiving client receives a
	  <function>wl_data_device.drop</function> event at which it is expected
	  to transfer the data using the
	  <function>wl_data_offer.receive</function> request.
	</para>
      </section>
    </section>
  </section>
</chapter>