Bug 1595601: Add documentation for the external handler service. r=Gijs

Differential Revision: https://phabricator.services.mozilla.com/D52578

--HG--
extra : moz-landing-system : lando

tools/docs/config.yml

@@ -12,6 +12,7 @@ categories:
         - modules/libpref
         - remote
         - services/common/services
+        - uriloader
     build_doc:
         - mach
         - tools/try

uriloader/docs/index.rst

@@ -0,0 +1,10 @@
+File Handling
+=============
+
+This covers how files requested for display are loaded.
+
+.. toctree::
+   :maxdepth: 2
+
+   uriloader
+   exthandler/docs/index

uriloader/docs/uriloader.rst

@@ -0,0 +1,46 @@
+.. _uri_loader_service:
+
+URI Loader Service
+==================
+
+As its name might suggest the URI loader service is responsible for loading URIs
+but it is also responsible for deciding how to handle that content, whether to
+display it as part of a DOM window or hand it off to something else.
+
+It is generally used when loading content for display to the user, normally from
+``nsDocShell`` for display as a webpage or ``nsObjectLoadingContent`` for display inside
+a webpage's ``<object>`` tag. The normal entrypoint is throught ``nsIURILoader::OpenURI``.
+
+The URI loader starts the load and registers an ``nsDocumentOpenInfo`` as a stream
+listener for the content. Once headers have been received `DispatchContent <https://searchfox.org/mozilla-central/search?q=nsDocumentOpenInfo%3A%3ADispatchContent&path=>`_
+then decides what to do with the content as it may need to be handled by something
+other than the caller. It uses a few criteria to decide this including:
+
+* Content-Type header.
+* Content-Disposition header.
+* Load flags.
+
+Part of this handling may include running the content through a registered stream
+converter to convert the content type from one to another. This is done through
+the `stream converter service <https://searchfox.org/mozilla-central/source/netwerk/streamconv>`_.
+When this happens a new ``nsDocumentOpenInfo`` is created to handle the new content
+in the same way as the current content.
+
+The rough flow goes as follows (though note that this can vary depending on the
+flags passed to the loader service):
+
+1. The caller may provide an ``nsIURIContentListener`` which can offer to handle
+   the content type or a content type that we can convert the original type to).
+   If so the load is passed off to the listener.
+2. Global instances of ``nsIURIContentListener`` can be registered with the URI
+   loader service so these are consulted in the same way.
+3. Global instances of ``nsIURIContentListener`` can be registered in the category
+   manager so these are consulted in the same way.
+4. Global instances of ``nsIContentHandler`` can be registered. If one agrees to
+   handle the content then the load is handed over to it.
+5. We attempt to convert the content to a different type.
+6. The load is handed over to the :ref:`External Helper App Service <external_helper_app_service>`.
+
+For the most part the process ends at step 1 because nsDocShell passes a ``nsDSURIContentListener``
+for the ``nsIURIContentListener`` consulted first and it accepts most of the
+`web content types <https://searchfox.org/mozilla-central/search?q=CONTENTDLF_CATEGORIES&redirect=false>`_.

uriloader/exthandler/docs/index.rst

@@ -0,0 +1,76 @@
+.. _external_helper_app_service:
+
+External Helper App Service
+===========================
+
+The external helper app service is responsible for deciding how to handle an
+attempt to load come content that cannot be loaded by the browser itself.
+
+Part of this involves using the Handler Service which manages the users
+preferences for what to do by default with different content.
+
+When a Link is Clicked
+----------------------
+
+When a link in a page is clicked (or a form submitted) ``nsDocShell`` tests
+whether the target protocol can be loaded by the browser itself, this is based
+on the preferences under ``network.protocol-handler``. When the browser cannot
+load the protocol it calls into ``nsExternalHelperAppService::LoadURI``.
+
+Some validation checks are performed but ultimateley we look for a registered
+protocol handler. First the OS is queried for an app registration for the
+protocol and then the handler server is asked to fill in any user settings from
+the internal database. If there were no settings from the handler service then
+some defaults are applied in ``nsExternalHelperAppService::SetProtocolHandlerDefaults``.
+
+If there is a default handler app chosen and the settings say to use it without
+asking then that happens. If not a dialog s shown asking the user what they
+want to do.
+
+During a Load
+-------------
+
+When content is already being loaded the :ref:`URI Loader Service <uri_loader_service>`
+determines whether the browser can handle the content or not. If not it calls
+into the external helper app server through ``nsExternalHelperAppService::DoContent``.
+
+The content type of the loading content is retrieved from the channel. A file
+extension is also generated using the Content-Disposition header or if the load
+is not a HTTP POST request the file extension is generated from the requested URL.
+
+We then query the MIME Service for an nsIMIMEInfo to find information about
+apps that can handle the content type or file extension based on OS and user
+settings, :ref:`see below <mime_service>` for further details. The result is
+used to create a ``nsExternalAppHandler`` which is then used as a stream listener
+for the content.
+
+The MIME info object contains settings that control whether to prompt the user
+before doing anything and what the default action should be. If we need to ask
+the user then a dialog is shown offering to let users cancel the load, save the
+content to disk or send it to a registered application handler.
+
+Assuming the load isn't canceled the content is streamed to disk using a background
+file saver with a target ``nsITransfer``. The ``nsITransfer`` is responsible for
+showing the download in the UI.
+
+If the user opted to open the file with an application then once the transfer is
+complete then ``nsIMIMEInfo::LaunchWithFile`` is used to
+`launch the application <https://searchfox.org/mozilla-central/search?q=nsIMIMEInfo%3A%3ALaunchWithFile&path=>`_.
+
+.. _mime_service:
+
+MIME Service
+------------
+
+The MIME service is responsible for getting an ``nsIMIMEInfo`` object for a
+content type or file extension:
+
+1. Fills out an ``nsIMIMEInfo`` based on OS provided information. This is platform
+   specific but should try to find the default application registered to handle
+   the content.
+2. Ask the handler service to fill out the ``nsIMIMEInfo`` with information held
+   in browser settings. This will not overwrite a any application found from
+   the OS.
+3. If one has not been found already then try to find a type description from
+   a `lookup table <https://searchfox.org/mozilla-central/search?q=extraMimeEntries[]&path=>`_
+   or just by appending " File" to the file extension.

uriloader/exthandler/moz.build

@@ -4,6 +4,8 @@
 # License, v. 2.0. If a copy of the MPL was not distributed with this
 # file, You can obtain one at http://mozilla.org/MPL/2.0/.
 
+SPHINX_TREES['docs'] = 'docs'
+
 TEST_DIRS += ['tests']
 
 XPIDL_SOURCES += [

uriloader/moz.build

@@ -4,6 +4,8 @@
 # License, v. 2.0. If a copy of the MPL was not distributed with this
 # file, You can obtain one at http://mozilla.org/MPL/2.0/.
 
+SPHINX_TREES['/uriloader'] = 'docs'
+
 with Files('**'):
     BUG_COMPONENT = ('Firefox', 'File Handling')