WebAssembly in a Hurry
Let’s take a quick tour of WebAssembly to get you familiar with foundations of Hippo.
Some programming languages, such as Go, Rust or C, are compiled languages. You select the operating system and CPU architecture where the code is going to execute, and it is turned into machine code suitable to run in that environment. If you support many operating systems and architectures, such as macos amd64, macos arm64 (aka M1), windows amd64, linux amd64, linux arm64, and so on, that you can spend a lot of time re-compiling your application for each combination.
Wouldn’t it be great if you could compile your application once and run it anywhere?
When you compile your application to WebAssembly, that is what happens! Write your application once in a language that supports compiling to WebAssembly (wasm), and then run on any target machine regardless of operating system or CPU architecture. This is possible because WebAssembly relies upon a WebAssembly runtime available on the target machine. The compiled output isn’t native machine code, and instead is byte code that is capable of being executed by a WebAssembly runtime.
A WebAssembly runtime handles executing a module in a sandbox, isolated from the host computer that is running the module. By default, modules cannot access host resources such as the filesystem or host network. The host (the person running the module) can choose to grant certain capabilities to a module. For example, one may allow a module to only load a specific file from the filesystem. WebAssembly makes sure that only that file can be accessed, and the module can’t “breakout” and access other files.
Given WebAssembly’s history, this model makes sense. WebAssembly was intended to execute in the web browser. And a browser should not inherently trust a random binary it downloaded from the internet. But that security model works well in other contexts. Standalone WebAssembly runtimes can be used to execute modules outside the browser.
If you are familiar with Docker containers, this will sound familiar. Containers can also be isolated. However they don’t execute in a true sandbox in the same way that WebAssembly does. Containers execute as processes directly on the host machine and use Linux functionality such as namespaces and cgroups to isolate the container. WebAssembly modules execute are compiled binaries that execute inside a WebAssembly virtual machine. The creator of Docker, Solomon Hykes, famously tweeted:
If WASM+WASI existed in 2008, we wouldn't have needed to created Docker. That's how important it is. Webassembly on the server is the future of computing. A standardized system interface was the missing link. Let's hope WASI is up to the task! https://t.co/wnXQg4kwa4— Solomon Hykes (@solomonstre) March 27, 2019
Some languages such as AssemblyScript, C, C#, Go, Rust, and Swift support compiling to target wasm, so that the resulting binary is ready to be executed by a WebAssembly runtime. There are many WebAssembly runtimes out there, such as [wasmtime] (used by Hippo) and [wasm3]. This means that you can often continue coding in your favorite language and with familiar developer tools and immediately target WebAssembly immediately.
WebAssembly System Interface (WASI)
Running in an isolated sandbox is great for security, but poses a challenge to developers when implementing traditional workloads. Questions immediately come up such as How do I serve static files associated with my application? or How do I communicate via sockets?
The WebAssembly format itself does not include a way for modules to access any resources outside of the host. To allow a module to access files, environment variables, and other system resources, an extension to WebAssembly was needed.
WebAssembly System Interface (WASI) defines an API through which WebAssembly modules can interact with the environment and the outside world, specifically with the filesystem and network. WASI defines a capabilities model in which the host (the person or process managing the runtime) can declare what the module is allowed to see or access. Thus, WASI makes it possible to safely and precisely identify which things a running module can access.
Not all languages that support WebAssembly also support WASI. We frequently use several languages that do support WASI: AssemblyScript, C, Rust, Grain, Zig, and Swift. As the WASI standard matures, we expect many other languages to add support.
WebAssembly Gateway Interface (WAGI)
Another common question that immediately comes up is Can I use WebAssembly to implement microservices? That is where WAGI comes in.
WebAssembly, even with WASI, does not provide access to networking primitives such as sockets. So it is not possible to compile a full web server to WebAssembly+WASI.
WebAssembly Gateway Interface (Wagi) takes a different, and well explored, approach. WAGI allows you to run WebAssembly WASI binaries as HTTP handlers. This pattern has been implemented for many other technologies, including PHP, Perl, Java Servlets, and ASP.net. But we took our inspiration from CGI.
Back in the day, web servers such as Apache, had support for writing HTTP request handlers in any language, as long as it adhered to the Common Gateway Interface (CGI). CGI was the precursor to what we now call Functions as a Service, or Serverless, and WAGI takes the next step so that you can define your functions in WebAssembly.
A Wagi-compliant server can be configured with mappings that define a
WebAssembly binary that should handle a path, such as all requests for
/calculator are routed to
calculator.wasm. The WebAssembly module is
executed for each incomming request, and its output is sent back as the
response to the client.
Wagi was created by Deis Labs, and is currently defined in a few documents that build off of CGI and highlight the differences:
The Wagi architecture is currently supported in two implementations: a standalone WAGI Server and an extension of .NET with WAGI.NET. Existing web servers could implement WAGI and support running WebAssembly modules too.