Every process to be interconnected via Mojo IPC is called a Mojo embedder and needs to embed the Embedder Development Kit (EDK) library. The EDK exposes the means for an embedder to physically connect one process to another using any supported native IPC primitive (e.g., a UNIX domain socket or Windows named pipe) on the host platform.
Details regarding where and how an application process actually embeds and configures the EDK are generaly hidden from the rest of the application code, and applications instead use the public System and Bindings APIs to get things done within processes that embed Mojo.
Once the EDK is initialized within a process, the public C System API is usable on any thread for the remainder of the process's lifetime. This is a lightweight API with a relatively small (and eventually stable) ABI. Typically this API is not used directly, but it is the foundation upon which all remaining upper layers are built. It exposes the fundamental capabilities to create and interact with various types of Mojo handles including message pipes, data pipes, and shared buffers.
There is a relatively small, higher-level system API for each supported language, built upon the low-level C API. Like the C API, direct usage of these system APIs is rare compared to the bindings APIs, but it is sometimes desirable or necessary.
The C++ System API provides a layer of C++ helper classes and functions to make safe System API usage easier: strongly-typed handle scopers, synchronous waiting operations, system handle wrapping and unwrapping helpers, common handle operations, and utilities for more easily watching handle state changes.
The Java System API provides helper classes for working with Mojo primitives, covering all basic functionality of the low-level C API.
Typically developers do not use raw message pipe I/O directly, but instead define some set of interfaces which are used to generate code that message pipe usage feel like a more idiomatic method-calling interface in the target language of choice. This is the bindings layer.
Interfaces are defined using the Mojom IDL, which can be fed to the bindings generator to generate code in various supported languages. Generated code manages serialization and deserialization of messages between interface clients and implementations, simplifying the code -- and ultimately hiding the message pipe -- on either side of an interface connection.
By far the most commonly used API defined by Mojo, the C++ Bindings API exposes a robust set of features for interacting with message pipes via generated C++ bindings code, including support for sets of related bindings endpoints, associated interfaces, nested sync IPC, versioning, bad-message reporting, arbitrary message filter injection, and convenient test facilities.
The Java Bindings API provides helper classes for working with Java code emitted by the bindings generator.
There are number of potentially decent answers to this question, but the deal-breaker is that a useful IPC mechanism must support transfer of native object handles (e.g. file descriptors) across process boundaries. Other non-new IPC things that do support this capability (e.g. D-Bus) have their own substantial deficiencies.
No. As an implementation detail, creating a message pipe is essentially generating two random numbers and stuffing them into a hash table, along with a few tiny heap allocations.
Yes! Nobody will mind. Create millions if you like. (OK but maybe don't.)
Compared to the old IPC in Chrome, making a Mojo call is about 1/3 faster and uses 1/3 fewer context switches. The full data is available here.
Yes, and message pipe usage is identical regardless of whether the pipe actually crosses a process boundary -- in fact this detail is intentionally obscured.
Message pipes which don't cross a process boundary are efficient: sent messages are never copied, and a write on one end will synchronously modify the message queue on the other end. When working with generated C++ bindings, for example, the net result is that an
InterfacePtr on one thread sending a message to a
Binding on another thread (or even the same thread) is effectively a
PostTask to the
TaskRunner with the added -- but often small -- costs of serialization, deserialization, validation, and some internal routing logic.
Please post questions to
email@example.com! The list is quite responsive.