NPAPI:Pepper: Difference between revisions

From MozillaWiki
Jump to navigation Jump to search
No edit summary
 
(110 intermediate revisions by 9 users not shown)
Line 1: Line 1:
= Status =
Mozilla is not interested in or working on Pepper at this time. See the Chrome [http://code.google.com/p/ppapi/ Pepper pages].
 
Under consideration.
 
== Background ==
 
Over the past few months a number of us have also been discussing some of the issues facing NPAPI as a platform independent browser plugin framework.  First, we feel there are some changes needed to NPAPI to keep up with recent developments in browser technology, such as out of process plugin execution.  Second, while NPAPI provides an extensive framework for writing plugins, many end up relying on operating system or browser platform specific features.  This is especially true of windowed plugins that implement 2D or 3D graphics, where almost the entirety of the plugin may consist of operating system specific graphics or event APIs. Third, both windowed and windowless plugins are difficult to composite with other layers correctly today, making it difficult for a page author to achieve the same look and feel across browsers and operating systems.  This proposal intends to be a starting point for addressing these issues, and consists of four topics.
 
# Giving a clear semantics to NPAPI across browsers that implement plugins in a separate process from the renderer/browser itself.
# Integrating plugin rendering with the browser's compositing process, allowing HTML overlays, etc., to work correctly with plugins.
# Defining events, 2D rasterization, and an initial take on 3D graphics access in a way that is common across operating systems and browsers.
# Determining which plugins are available to a browser without loading the plugin.  Although a much smaller issue, it is currently one of the more platform specific portions of NPAPI and one that lends itself to a relatively simple solution.
 
A somewhat related issue is how to extend NPAPI to provide platform independent access to new hardware or software features such as webcams, microphones, etc.  While certainly important, this is deferred to future discussions.
 
== Out of process plugins ==
 
The current NPAPI model defines the threading model within one process such that whatever thread invokes NPP_New is the only thread that may invoke NPN_* functions.  Several browsers either currently support or are considering moving NPAPI plugins to a separate process from the renderer process (responsible for rendering a page 's content).  With these processes distinct, we have the following: 
 
<pre>
R = renderer process
Tr = thread in R that causes NPP_New to be invoked
P = plugin process
Tp = thread in P that invokes NPP_New in response to the renderer
</pre>
 
=== Thread model for NPN_* calls ===
We propose that NPN_* calls, with only one exception, can only be made from Tp.  This allows us to maintain an important invariant, namely that if Tp is executing any NPP code, then Tr is blocked.  This eliminates the possibility of simultaneous JavaScript and plugin activities.
 
=== NPN_PluginThreadAsyncCall ===
 
The one exception to NPN_* calls is NPN_PluginThreadAsyncCall.
 
<pre>
void    NP_LOADDS NPN_PluginThreadAsyncCall(NPP instance,
                                            void (*func) (void *),
                                            void *userData);
</pre>
 
The semantics of NPN_PluginThreadAsyncCall remain fundamentally the same, just with respect to P:
* It can be called from any thread in P.
* Invoking it eventually causes an NPP callback on Tp to func, passing userData.
 
We make a couple of other implementation comments that make this more usable and help avoid possible deadlocks:
* If two callbacks are requested by the same thread, they should be invoked in the same order they were requested.
* The implementation of NPN_PluginThreadAsyncCall may not call func directly, even if called on Tp, because we need to guarantee the state of Tr and this risks deadlock.
 
We observe that implementing NPN_PluginThreadAsyncCall this way results in additional call/callback/call latency for NPN_* calls.  We believe this is reasonable to provide a well-defined interaction, but realize we may need to address the efficiency of this mechanism.
 
== Plugins and rendering ==
 
Currently plugins typically display 2D graphics in one of two modes: windowless or windowed.  In the current state of affairs windowless plugins are typically given an RGB surface containing the contents of lower layers and they composite themselves.  This constrains the use of transparency and complicates plugins that are not the top or bottom layer.  Furthermore, basically all access to 3D accelerated graphics are through the latter.  Windowed plugins provide a number of challenges to plugin and browser writers.  First, windowed plugins are essentially given a handle to an operating system native window, from which all interaction by the plugin is completely platform-specific.  Second, native windows make it very difficult for the browser to do CSS overlays, etc.  Third, although NPAPI provides the NPP_HandleEvent API, this is not well-defined with respect to native windowing system events.
 
To these ends we propose, for both 2D and 3D:
* Only windowless plugins should be supported.
* Compositing should be done by outside the plugin itself, which should only produce an RGBA (2D) or a texture (3D).
* No native windowing system events should be delivered to plugins.
 
This proposal is clearer at this moment regarding 2D graphics. We expect we may need to pass some state (2d vs. 3d, hardware vs. raster, etc.) in through other calls and/or with some extensions to the data structures we propose below.
 
=== Negotiating the drawing model ===
 
A plugin may call NPN_GetValue() with the following NPNVariable to query the browser whether it supports the Pepper drawing model:
 
<pre>
/* TRUE if the browser supports the PepperGraphics drawing model */
NPNVsupportsPepperGraphicsBool = xxxx
</pre>
 
Once the plugin finds a supported drawing model, it calls NPN_SetValue() to tell the browser which drawing model it will use. We're using the NPNVariable used for Mac drawing model for this:
 
<pre>
NPNVpluginDrawingModel = 1000 /* The NPDrawingModel specified by the plugin */
</pre>
 
We will extend the NPDrawingModel enumeration proposed for Mac with another model:
<pre>
typedef enum {
    ...
    NPDrawingModelPepperGraphics = yyy,
} NPDrawingModel;
</pre>
 
=== An Example ===
 
A video game looks abstractly like
<pre>
  ...
  while (true) {
    HandleEvents()
    Simulate(gettimeofday())
    Draw()
  }
  ...
</pre>
 
We propose that we continue to set graphics attributes through the existing NPN_GetValue/NPN_SetValue interfaces with some new variables TBD.
 
There are four APIs that would be involved in supporting this from the browser.
 
=== NPP_SetWindow and Windowless Plugins ===
 
Whereas currently NPP_SetWindow passes both some geometry information and a handle to a native window (for windowed plugins), in our proposal, it only conveys geometry and format information.  For example.
 
<pre>
typedef struct _NPWindow {
  void*    window;  // UNUSED
  uint32_t x;        // Position of plugin top left corner relative to top left
  uint32_t y;        // of renderer page. Y increases going down.
  uint32_t width;    // maximum window size
  uint32_t height;
  NPRect  clipRect; // UNUSED
  void*    ws_info;  // Pixel format of the raster backing store. A NPRasterWindowInfo struct.
  NPWindowType type; // UNUSED
} NPWindow;
 
enum NPPixelFormat {
  NP_RGBA_8888,
  NP_RGB_565,
  // TBD: other pixel formats
};
 
typedef struct _NPRasterWindowInfo {
  NPPixelFormat format;
} NPRasterWindowInfo;
 
NPError NP_LOADDS NPP_SetWindow(NPP instance,
                                NPWindow* window);
 
</pre>
 
=== Getting a Device Context to Draw Into ===
<pre>
  PARAMS (details remain to be worked out, but at least):
    - rectangle plugin will draw into
    - discard bit (don't bother to initialize the memory in that rectangle)
  2D DC:
    - x, y, width, height
    - visible rectangles[] // serve as hints
    - stride (byte/pixel difference between (x,y) and (x,y+1))
    - x, y for byte zero of the buffer
    - void* buffer // size is as requested by PARAMS regardless of visible rectangles.
 
  void NPN_GetDeviceContext(NPP instance, PARAMS, NPDeviceContext** context)
</pre>
 
* Allows the plugin to request that the renderer create a context to draw into and call the plugin back.
* On return, the renderer informs the plugin where it can draw into (in *context).
 
=== Publishing the Context to the Renderer ===
<pre>
  void NPN_PublishDeviceContext(NPP instance, NPDeviceContext* context)
</pre>
 
* The plugin informs the renderer that it has drawn into context and that the renderer may composite from it.
* After this call the plugin may no longer draw into the context (it may be deallocated, etc.)
 
=== Notification a Context was Displayed ===
 
Video players, video games, and other timing sensitive plugins typically adjust their simulation to adapt to the speed at which frames are placed on the screen.  A call from the renderer indicating when the bits from a context have reached the screen is needed.
 
<pre>
  void NPP_DeviceContextDisplayed(NPP instance)
</pre>
 
* The renderer informs the plugin that the last published device context was displayed on the screen.
* For performance we will most likely want to be able to call NPN_GetDeviceContext before the NPP_DeviceContextDisplayed call from the previous NPN_PublishDeviceContext has arrived.
* This API could also perhaps be done through an NPEvent of the appropriate type.
 
=== Optimizations and Legacy APIs ===
 
There are two possible performance opportunities that should be noted.
First, given the threading model we have defined, it may be desirable to allow NPN_GetDeviceContext and NPN_PublishDeviceContext to be called from other plugin threads than Tp.  This would eliminate the need to do the associated NPN_PluginThreadAsyncCall, and avoid the context switch penalties that might be associated.
 
Second, in order to avoid drawing an entire region on the screen when only a small portion is updated, it may be useful for the plugin to use existing APIs to mark only a small region as needing to be updated by the next NPN_PublishDeviceContext.
 
<pre>
  void NPN_InvalidateRect(NPP instance, NPRect* invalidRect)
</pre>
* The plugin informs the renderer that it would like to replace the contents of the specified rectangle.
 
== Event Handling ==
 
Plugins should receive all their events from NPP_HandleEvent.  We believe that standardizing the event types and payloads based on, for example, DOM events, would be sufficient to provide browser and OS independent events.  If not, the goal can be accomplished by possibly extending the set of event types or data on the event structures to convey more information.
 
== Plugin Registration ==
 
The current process of determining which plugins are available to the browser typically involves loading various shared libraries and querying them for types and extensions addressed by the plugin.  We propose that more efficient method would be to add a special section to the shared library for the plugin.
 
The section, tentatively named .npapidesc, contains a sequence of strings of the form:
 
<pre>
mimetype:[extension][,extension]*:description;
</pre>
 
This enables browsers to scan the plugins without loading them, which becomes more difficult with out-of-process plugins and sandbox protection.

Latest revision as of 22:31, 26 May 2011

Mozilla is not interested in or working on Pepper at this time. See the Chrome Pepper pages.

Retrieved from "https://wiki.allizom.org/index.php?title=NPAPI:Pepper&oldid=313199"