The configuration file config.xml =========== The widgets are described by the W3C's technical recommendations [Packaged Web Apps (Widgets)][widgets] and [XML Digital Signatures for Widgets][widgets-digsig] that specifies the configuration file **config.xml**. Overview -------- The file **config.xml** describes important data of the application to the framework: - the unique identifier of the application - the name of the application - the type of the application - the icon of the application - the permissions linked to the application - the services and dependancies of the application The file MUST be at the root of the widget and MUST be case sensitively name ***config.xml***. The file **config.xml** is a XML file described by the document [widgets]. Here is the example of the config file for the QML application SmartHome. ```xml <?xml version="1.0" encoding="UTF-8"?> <widget xmlns="http://www.w3.org/ns/widgets" id="smarthome" version="0.1"> <name>SmartHome</name> <icon src="smarthome.png"/> <content src="qml/smarthome/smarthome.qml" type="text/vnd.qt.qml"/> <description>This is the Smarthome QML demo application. It shows some user interfaces for controlling an automated house. The user interface is completely done with QML.</description> <author>Qt team</author> <license>GPL</license> </widget> ``` The most important items are: - **<widget id="......"\>**: gives the id of the widget. It must be unique. - **<widget version="......"\>**: gives the version of the widget - **<icon src="..."\>**: gives a path to the icon of the application (can be repeated with different sizes) - **<content src="..." type="..."\>**: this indicates the entry point and its type. Standard elements of "config.xml" --------------------------------- ### The element widget #### the attribute id of widget The attribute *id* is mandatory (for version 2.x, blowfish) and must be unique. Values for *id* are any non empty string containing only latin letters, arabic digits, and the three characters '.' (dot), '-' (dash) and '\_' (underscore). Authors can use a mnemonic id or can pick a unique id using command **uuid** or **uuidgen**. ### the attribute version of widget The attribute *version* is mandatory (for version 2.x, blowfish). Values for *version* are any non empty string containing only latin letters, arabic digits, and the three characters '.' (dot), '-' (dash) and '\_' (underscore). Version values are dot separated fields MAJOR.MINOR.REVISION. Such version would preferabily follow guidelines of [semantic versionning][semantic-version]. ### The element content The element *content* is mandatory (for version 2.x, blowfish) and must designate a file (subject to localisation) with its attribute *src*. The content designed depends on its type. See below for the known types. ### The element icon The element *icon* is mandatory (for version 2.x, blowfish) and must be unique. It must designate an image file with its attribute *src*. AGL features ------------ The AGL framework uses the feature tag for specifying security and binding requirement of the widget. Since the migration of the framework to leverage systemd power, the features are of important use to: - declare more than just an application - declare the expected dependencies - declare the expected permissions - declare the exported apis The specification of [widgets][widgets] is intentded to describe only one application. In the present case, we expect to describe more than just an application. For example, a publisher could provide a widget containing a service, an application for tuning that service, an application that leverage the service. Here, the term of service means a background application that runs without IHM and whose public api can be accessed by other applications. So the features are used to describe each of the possible units of widgets. The "standard" unit in the meaning of [widgets][widgets] is called the "main" unit. ### feature name="urn:AGL:widget:required-api" List of the api required by the widget. Each required api must be explicited using a <param> entry. Example: ```xml <feature name="urn:AGL:widget:required-api"> <param name="#target" value="main" />> <param name="gps" value="auto" /> <param name="afm-main" value="link" /> </feature> ``` This will be *virtually* translated for mustaches to the JSON ```json "required-api": [ { "name": "gps", "value": "auto" }, { "name": "afm-main", "value": "link" } ] ``` #### param name="#target" OPTIONAL Declares the name of the unit requiring the listed apis. Only one instance of the param "#target" is allowed. When there is not instance of this param, it behave as if the target main was specified. #### param name=[required api name] The name is the name of the required API. The value describes how to connect to the required api. It is either: - local: The binding is a local shared object. In that case, the name is the relative path of the shared object to be loaded. - auto: The framework set automatically the kind of the connection to the API - ws: The framework connect using internal websockets - dbus: The framework connect using internal dbus - link: The framework connect in memory by dinamically linking - cloud: [PROPOSAL - NOT IMPLEMENTED] The framework connect externally using websock. In that case, the name includes data to access the service. Example: `<param name="log:https://oic@agl.iot.bzh/cloud/log" value="cloud" />` ### feature name="urn:AGL:widget:required-permission" List of the permissions required by the unit. Each required permission must be explicited using a <param> entry. Example: ```xml <feature name="urn:AGL:widget:required-permission"> <param name="#target" value="geoloc" /> <param name="urn:AGL:permission:real-time" value="required" /> <param name="urn:AGL:permission:syscall:*" value="required" /> </feature> ``` This will be *virtually* translated for mustaches to the JSON ```json "required-permission":{ "urn:AGL:permission:real-time":{ "name":"urn:AGL:permission:real-time", "value":"required" }, "urn:AGL:permission:syscall:*":{ "name":"urn:AGL:permission:syscall:*", "value":"required" } } ``` #### param name="#target" OPTIONAL Declares the name of the unit requiring the listed permissions. Only one instance of the param "#target" is allowed. When there is not instance of this param, it behave as if the target main was specified. #### param name=[required permission name] The value is either: - required: the permission is mandatorily needed except if the feature isn't required (required="false") and in that case it is optional. - optional: the permission is optional ### feature name="urn:AGL:widget:provided-unit" This feature is made for declaring new units for the widget. Using this feature, a software publisher can provide more than one application in the same widget. Example: ```xml <feature name="urn:AGL:widget:provided-unit"> <param name="#target" value="geoloc" /> <param name="description" value="binding of name geoloc" /> <param name="content.src" value="index.html" /> <param name="content.type" value="application/vnd.agl.service" /> </feature> ``` This will be *virtually* translated for mustaches to the JSON ```json { "#target":"geoloc", "description":"binding of name geoloc", "content":{ "src":"index.html", "type":"application\/vnd.agl.service" }, ... } ``` #### param name="#target" REQUIRED Declares the name of the unit. The default unit, the unit of the main of the widget, has the name "main". The value given here must be unique within the widget file. It will be used in other places of the widget config.xml file to designate the unit. Only one instance of the param "#target" is allowed. The value can't be "main". #### param name="content.type" REQUIRED The mimetype of the provided unit. #### param name="content.src" A path to the #### other parameters The items that can be set for the main unit can also be set using the params if needed. - description - name.content - name.short - ... ### feature name="urn:AGL:widget:provided-api" Use this feature for exporting one or more API of a unit to other widgets of the platform. This feature is an important feature of the framework. Example: ```xml <feature name="urn:AGL:widget:provided-api"> <param name="#target" value="geoloc" /> <param name="geoloc" value="auto" /> <param name="moonloc" value="auto" /> </feature> ``` This will be *virtually* translated for mustaches to the JSON ```json "provided-api":[ { "name":"geoloc", "value":"auto" }, { "name":"moonloc", "value":"auto" } ], ``` #### param name="#target" OPTIONAL Declares the name of the unit exporting the listed apis. Only one instance of the param "#target" is allowed. When there is not instance of this param, it behave as if the target main was specified. #### param name=[name of exported api] The name give the name of the api that is exported. The value is one of the following values: - ws: export the api using UNIX websocket - dbus: export the API using dbus - auto: export the api using the default method(s). Known content types ------------------- The configuration file ***/etc/afm/afm-unit.conf*** defines how to create systemd units for widgets. Known types for the type of content are: - ***text/html***: HTML application, content.src designates the home page of the application - ***application/vnd.agl.native*** AGL compatible native, content.src designates the relative path of the binary. - ***application/vnd.agl.service***: AGL service, content.src is not used. - ***application/x-executable***: Native application, content.src designates the relative path of the binary. For such application, only security setup is made. Adding more types is easy, it just need to edit the configuration file ***afm-unit.conf***. ### Older content type currently not supported at the moment. This types were defined previously when the framework was not leveraging systemd. The transition to systemd let these types out at the moment. - ***application/vnd.agl.url*** - ***text/vnd.qt.qml***, ***application/vnd.agl.qml*** - ***application/vnd.agl.qml.hybrid*** - ***application/vnd.agl.html.hybrid*** The configuration file afm-unit.conf ==================================== The integration of the framework with systemd mainly consists of creating the systemd unit files corresponding to the need and requirements of the installed widgets. This configuration file named `afm-unit.conf` installed on the system wiht the path `/etc/afm/afm-unit.conf` describes how to generate all units from the *config.xml* configuration files of widgets. The description uses an extended version of the templating formalism of [mustache][] to describes all the units. Let present how it works using the following diagram that describes graphically the workflow of creating the unit files for systemd `afm-unit.conf` from the configuration file of the widget `config.xml`: ![make-units][make-units] In a first step, and because [mustache][] is intended to work on JSON representations, the configuration file is translated to an internal JSON representation. This representation is shown along the examples of the documentation of the config files of widgets. In a second step, the mustache template `afm-unit.conf` is instanciated using the C library [mustach][] that follows the rules of [mustache][mustache] and with all its available extensions: - use of colon (:) for explicit substitution - test of values with = or =! In a third step, the result of instanciating `afm-unit.conf` for the widget is splited in units. To achieve that goal, the lines containing specific directives are searched. Any directive occupy one full line. The directives are: - %nl Produce an empty line at the end - %begin systemd-unit - %end systemd-unit Delimit the produced unit, its begin and its end - %systemd-unit user - %systemd-unit system Tells the kind of unit (user/system) - %systemd-unit service NAME - %systemd-unit socket NAME Gives the name and type (service or socket) of the unit. The extension is automatically computed from the type and must not be set in the name. - %systemd-unit wanted-by NAME Tells to install a link to the unit in the wants of NAME Then the computed units are then written to the filesystem and inserted in systemd. The generated unit files will contain variables for internal use of the framework. These variables are starting with `X-AFM-`. The variables starting with `X-AFM-` but not with `X-AFM--` are the public variables. These variables will be returned by the framework as the details of an application (see **afm-util detail ...**). Variables starting with `X-AFM--` are private to the framework. By example, the variable `X-AFM--http-port` is used to record the allocated port for applications. [mustach]: https://gitlab.com/jobol/mustach "basic C implementation of mustache" [mustache]: http://mustache.github.io/mustache.5.html "mustache - Logic-less templates" [make-units]: pictures/make-units.svg [widgets]: http://www.w3.org/TR/widgets "Packaged Web Apps" [widgets-digsig]: http://www.w3.org/TR/widgets-digsig "XML Digital Signatures for Widgets" [libxml2]: http://xmlsoft.org/html/index.html "libxml2" [app-manifest]: http://www.w3.org/TR/appmanifest "Web App Manifest" [meta-intel]: https://github.com/01org/meta-intel-iot-security "A collection of layers providing security technologies" [widgets]: http://www.w3.org/TR/widgets "Packaged Web Apps" [widgets-digsig]: http://www.w3.org/TR/widgets-digsig "XML Digital Signatures for Widgets" [libxml2]: http://xmlsoft.org/html/index.html "libxml2" [openssl]: https://www.openssl.org "OpenSSL" [xmlsec]: https://www.aleksey.com/xmlsec "XMLSec" [json-c]: https://github.com/json-c/json-c "JSON-c" [d-bus]: http://www.freedesktop.org/wiki/Software/dbus "D-Bus" [libzip]: http://www.nih.at/libzip "libzip" [cmake]: https://cmake.org "CMake" [security-manager]: https://wiki.tizen.org/wiki/Security/Tizen_3.X_Security_Manager "Security-Manager" [app-manifest]: http://www.w3.org/TR/appmanifest "Web App Manifest" [tizen-security]: https://wiki.tizen.org/wiki/Security "Tizen security home page" [tizen-secu-3]: https://wiki.tizen.org/wiki/Security/Tizen_3.X_Overview "Tizen 3 security overview" [semantic-version]: http://semver.org/ "Semantic versionning"