A manifesto for cloud-oriented programming.
Don't get me wrong, I love the cloud! It has empowered me to build amazing things, and completely changed the way I use software to innovate and solve problems.
It's the "new computer", the ultimate computer, the "computerless computer". It can elastically scale, it's always up, it exists everywhere, it can do anything. It's boundless. It's definitely here to stay.
But holy crap, there is no way this is how we are going to be building applications for the cloud in the next decade. As the cloud evolved from "I don't want servers under my desk" to "my app needs 30 different managed services to perform its tasks", we kind of lost track of what a great developer experience looks like.
Building applications for the cloud sometimes feels like spilling my kids' bag of unused Lego blocks all over the living room floor and trying to build a castle. After going through torn up play cards, scary Barbie-doll heads, and leaking dead batteries, you read the instructions the millionth time, only to realize you ended up building basically the same thing you've built last time.
Sorting Lego blocks is fun! It passes the time with the kiddos. It even feeds my OCD… But hell, this is not how I want to build professional software!
Let me try to describe what I and my developer friends are struggling with.
I want to focus on creating value for my users
When I build professional software, I want most of my time to be spent within the functional domain of my application, instead of non-functional mechanics of the platform I use.
It doesn't make sense that every time I want to execute code inside an AWS Lambda function, I have to understand that it needs to be bundled with tree-shaken dependencies, uploaded as a zip file to S3 and deployed through Terraform. Or that in order to be able to publish a message to SNS, my IAM policy must have a statement that allows the sns:Publish action on the topic's ARN. And does every developer need to understand what ARNs are at all?
All that stuff doesn't have anything to do with the value I am trying to create for my users. It's pure mechanics. Can we get rid of it?
I want to be independent
One of the most frustrating and flow killing situations for me as a developer is when I have to stop and wait for someone or something in order to continue.
It's like gliding happily in the air, enjoying the view, beautiful music in the background, and suddenly, BAM! A concrete wall.
This concrete wall takes many shapes and sizes when you build applications for the cloud. It's the DevOps person with an endless ticket queue; it's the IAM policy that needs to be updated; it's the deployment failure that only the external part-time consultant knows how to debug; It's the endless backlog of missing knobs and APIs in the internal platform that we hoped will change everything.
These barriers are frustrating because they force me to switch context, to apply "temporary" security policies and to invent ugly hacks that I don't want to talk about. It's a broken world.
I want to be independent. I want to be able to get things done, to stay in the flow. I want to improve the world one commit at a time, and move on to the next thing after I am finished. I want that dopamine rush of completing a task, not the shameful feeling of yet another unfinished thread.
I want instant feedback
I said I want independence, but don't mistake that for a belief that I write perfect code. Which is why I want to write code with a pencil, not with a pen.
Some developers can spend a full day coding without even invoking their compiler, and at the end of the day, they compile and deploy, and it just works.
I admire them, but I am not that type of developer. No sir. For me it's about iterations, iterations, iterations. I start small, sketch with a light pencil, take a look, erase a bunch of stuff, draw a thicker line, take a step back, squint, draw more and erase more, and take another look, rinse and repeat.
This is why, for me, the single most important thing is iteration speed. The sooner I can run my application and test it, the faster I can go back and iterate. This is where my flow is.
When I started programming, I used Borland C++. It used to take about 100ms to compile and run a program on an IBM PC AT machine (TURBO ON). An average iteration cycle in the cloud takes minutes. Minutes! Sometimes dozens of minutes!
Here's how an iteration looks like in the cloud today: I make a change to my code; then I need to compile it; deploy it to my test account; find my way around the management console to actually trigger it; wait for it to run and go search for the
logs on another service. Then I realize there is an error response that tells me that I'm stupid, because how come I didn't know that I have to pass in Accept-Content: application/json, because otherwise I get some weird result called "XML" that I have no idea what to do with (just kidding, XML is great, no really). Now all over again...
So "write unit tests", you say, in a patronizing attempt to justify the current reality. "Great developers write unit tests". OK! So now I need to take my code, which makes about 20 external API calls, and somehow mock out the API responses by copying and pasting them from outdated documentation, only to figure out that my requests are rejected because I am missing some implicit action in my IAM security statement. We've all been there.
To be honest, give me the developer experience of the 90s. I want to make a change, and I want to be able to test this change either interactively or through a unit test within milliseconds, and I want to do this while sitting in an airplane with no WiFi, okay? (we didn't have WiFi in the 90s).
So this is just a rant?
Hell no! I am a programmer. I sometimes feel like I've been writing software since birth. I've been doing it in socially perilous times, when being a computer geek was not cool.
What I have always loved about being a developer is that if I was not happy with my tools, I could make my own. Building tools is in our DNA, after all - humans have been building tools for over a million years.
And I am not happy with my tools.
In April 2022, I joined forces with Shai Ber, a good friend and a former Microsoft colleague, and we founded [Wing] with the mission to unlock the cloud for developers. We've assembled an incredible crew of beautiful geeks who share our passion for developer experience and open-source and started our journey to empower developers (i.e. ourselves) to solve these fundamental problems.
Compilers to the rescue
So how are we going to solve all of these problems at once?
We are building a programming language for the cloud.
"A programming language!?," you ask. "Doesn't the world have enough programming languages?," "Isn't it really hard to write a compiler?," "What are the chances that developers will want to learn a whole new language?," "Why can't you hack into an existing language toolchain, squint your eyes tight enough and call it a day?"
I am not one to build programming languages on a whim. In fact, I've spent the last five years building the AWS CDK, which is a multi-language library that addresses some of the challenges I am talking about by allowing developers to define cloud infrastructure using their favorite programming language.
To "meet developers where they are" is a beautiful tenet of AWS, and of the CDK, and inspired us to create awesome technology such as JSII and constructs.
But sometimes, "where they are" is not a good enough model for creating the desired experience.
Defining infrastructure with code does enable us to create a higher-level of abstraction, but as long as my application code needs to interact with this infrastructure, the abstraction becomes too leaky. I'm yanked back down to having to understand more than I need to, and I have to be an expert in things like IAM, VPC, ALB, EBS and basically more TLAs than I would ever want to keep in my head.
The languages we use today are all designed around the idea that the computer is a single machine. They've reached the point in which they are able to offer us solid abstractions over these machines. They abstract away the CPU, memory, file system, process management and networking. As a developer, I don't have to care how a file is laid out on disk, or even how much memory I need for my hash map. I simply write readFile() or new Dictionary() and go about my day. Yes, it's not a bad idea for me to have some sense of what's happening under the hood, but I am not forced to.
Most of these languages also offer me type-safety. When I call a function with the wrong number of arguments, I get yelled at by my compiler. I don't have to wait until my application is running only to realize I forget an argument, or passed in the wrong type.
In the cloud, I'm on my own. Every time my code needs to interact with a cloud resource or a service - and that's happening more and more as the industry evolves - I have to leave the comfort and safety of my programming language. I must jump outside the boundaries of the machine and into the wild wild west of the internet, and my compiler is none the wiser.
And suddenly, it's almost painfully obvious where all the pain came from. Cloud applications today are simply a patchwork of disconnected pieces. I have a compiler for my infrastructure, another for my functions, another for my containers, another for my CI/CD pipelines. Each one takes its job super seriously, and keeps me safe and happy inside each of these machines, but my application is not running on a single machine anymore, my application is running on the cloud.
The cloud is the computer.
Wing, a cloud-oriented programming language
When new programming paradigms emerge, it takes languages time to catch up. I used to love building object-oriented code in C, but it was a leaky abstraction. I had to understand how objects are laid out in memory, how V-tables work, and remember to pass the object as the first argument for each function. When programming languages started to support object-oriented concepts as first-class citizens, this paradigm was democratized, and today most developers don't even know what V-tables are, and the world keeps spinning.
Wing, or winglang if you want to be cute about it, has all the good stuff you would expect from a modern, object oriented, strongly-typed and general-purpose language, but it also includes a few additional primitives designed to support the distributed and service-based nature of the cloud as first-class citizens.
Check it out
We have been working on Wing for almost a year now, and I am excited to invite you to check it out and let me know what you think. While still in Alpha and not yet ready for production use, it's already possible to build some real applications with it.
Top comments (16)
All abstractions can break when things go wrong but that doesn’t mean they are not powerful and valuable. Abstractions are the essence of software as they enable developers to deliver more value to our users faster. When things don’t work, we gotta open the hood. That’s been like that with software since the dawn of computers ❤️
I love the compiler approach! We're also building a language (wasp-lang.dev/), so I'm aware of the challenges that come with it? Have you implemented your LSP yet?
Wasp looks sick! Yes, Wing's toolchain is designed with an LSP-first approach so it reuses the same code for parsing and type checking with the core compiler. :-)
nice, the same!
Have you guys heard about pulumi. If not, welcome to Pulumi, much user friendly, code with your favorite programming language and target your favorite cloud of your choice. They are the best one to refer. I love them :)
Advice - Do not reinvent the wheel. Humans will be dead, AI's will take over LOL
Love this manifesto! Great read and converted me ⭐️
Man.. really enjoyed that one. Keep dem coming :D
Very good read, thanks!
Thank you
Great stuff!
What a great write-up @eladb!
superb, resonates with my thoughts about working with cloud!
The idea that you to have to push to the cloud with each iteration is bonkers. I haven't worked like that since 2008.
Since then, I've got all the servers I need (minus occasionally some 3rd party APIs) running locally. At each iteration, I point curl at localhost and test it myself.
I have worked from a plane with no Wi-Fi recently. When my Internet goes out at home, I keep working.
Some comments may only be visible to logged-in visitors. Sign in to view all comments.