Trisiel Progress: Hello, World!
Published on 12/08/2019, 1398 words, 6 minutes to read
I have been working off and on over the years and have finally created the base of a functions as a service backend for WebAssembly code. I'm code-naming this wasmcloud. Trisiel is a pre-alpha prototype and is currently very much work in progress. However, it's far enough along that I would like to explain what I have been doing for the last few years and what it's all built up to.
Here is a high level view of all of the parts that make up wasmcloud and how they correlate:
Land: The Beginning
A little bit after I found WebAssembly I started to play with it. It seemed like it was too good to be true. A completely free and open source VM format that would run on almost any platform? Sounds like the kind of black magick witchcraft you hear about on Star Trek.
However, I kept at it and continued experimenting. I eventually came up with Land. This was a very simple thing and was really used to help me invent Dagger.
Dagger was an attempt at an incredible amount of minimalism. I based it on an extreme interpretation of the Unix philosophy (everything is a file -> everything is a bytestream) combined with some Plan 9 for flavor. It had only 5 system calls:
open()- opens a stream by URL, returning a stream descriptor
close()- closes a stream descriptor
read()- reads from a stream
write()- writes to a stream
flush()- flushes intermediate data and turns async behavior into syncronous behavior
And yet this was enough to implement a HTTP client.
The core guiding idea was that a cloud-native OS API should expose internet resources as easily as it exposes native resources. It should be as easy to use WebSockets as it is to use normal sockets. Additionally, all of the details should be abstracted away from the WebAssembly module. DNS resolution is not its job. TLS configuration is not its job. Its job is to run your code. Everything else should just be provided by the system.
And this worked for several months as I learned WebAssembly and started to experiment with bigger and better things.
Olin: Phase 2
Land taught me a lot. I started to quickly run into the limits of Dagger though. I ended up needing calls like non-cryptographic entropy, environment variables, command-line arguments and getting the current time. After doing some research (and trying/failing to implement my own such API based on newlib) I found a library and specification called CommonWA. This claimed to offer a lot of what I was looking for. Namely URLs as filenames and all of the host interop support I could hope for. I named this platform Olin, or the One Language Intelligent Network.
However the specification was somewhat dead. The author of it had largely moved on to more ferrous pastures and I became one of the few users of it. I ended up forking the specification and implementing my view of what it should be.
I ended up implementing a Rust implementation of the guest -> host API for the Webassembly side of things. I forked some of the existing Rust code for this and gradually started adding more and more things. The test harness is the biggest wasm program I've written for a while. Seriously, there's a lot going on there. It tests every single function exposed in the CWA spec as well as all of the schemes I had implemented.
Over time I ended up testing Olin in more and more places and on more and more hardware. As a side effect of all of this being pure go, it was very easy to cross compile for PowerPC, 32 bit arm (including a $9 arm board that lives under my desk) and even other targets that gccgo supports. I even ended up porting part of TempleOS to Olin as a proof of concept, but have more plans in the future for porting other parts of its kernel as a way to help people understand low-level operating system development.
I've even written a few blogposts about Olin:
But, this was great for running stuff interactively and via the command line. It left me wanting more. I wanted to have that mythical functions as a service backend that I've been dreaming of. So, I created Trisiel.
As an interlude, I also created the h programming language during this time as a satirical parody of V. This ended up helping me test a lot of the core functionality that I had built up with Olin. Here's an example of a program in h:
And this compiles to:
(module (import "h" "h" (func $h (param i32))) (func $h_main (local i32 i32 i32) (local.set 0 (i32.const 10)) (local.set 1 (i32.const 104)) (local.set 2 (i32.const 39)) (call $h (get_local 1)) (call $h (get_local 0)) ) (export "h" (func $h_main)) )
This ends up printing:
I think this is the smallest (if not one of the smallest) quine generator in the world. I even got this program running on bare metal:
Trisiel is the culmination of all of this work. The goal of wasmcloud is to create a functions as a service backend for running people's code in an isolated server-side environment.
Users can use the
wasmcloud command line tool to do everything at the moment:
$ wasmcloud Usage: wasmcloud <flags> <subcommand> <subcommand args> Subcommands: commands list all command names flags describe all known top-level flags help describe subcommands and their syntax Subcommands for api: login logs into wasmcloud whoami show information about currently logged in user Subcommands for handlers: create create a new handler logs shows logs for a handler Subcommands for utils: namegen show information about currently logged in user run run a webassembly file with the same environment as production servers Top-level flags (use "wasmcloud flags" for a full list): -api-server=http://wasmcloud.kahless.cetacean.club:3002: default API server -config=/home/cadey/.wasmc.json: default config location
This tool lets you do a few basic things:
- Authenticate with the Trisiel server
- Create handlers from WebAssembly files that meet the CommonWA API as realized by Olin
- Get logs for individual handler invocations
- Run WebAssembly modules locally like they would get run on Trisiel
Nearly all of the complexity is abstracted away from users as much as possible.
In the future I hope to do the following things:
- Support updating handlers to new versions of the code
- Support live-streaming of logs
- Support handler deletion
- Support bulk queue export
- Support wasi for easier interoperability
- Support more resource types such as websockets
- Investigate porting the wasmcloud executor to Rust
- Documentation/a book on how to use wasmcloud
- Create an easier way to create accounts that can make handlers
- Deploy to production somewhere
Every single one of these people was immeasurably helpful in this research over the years.
And many more I can't remember because it's been so many.
If you want to support my work, please do so via Patreon. It really means a lot to me and helps to keep the dream alive!
Facts and circumstances may have changed since publication. Please contact me before jumping to conclusions if something seems wrong or unclear.
Tags: wasm, faas