Intro and Motivation

Many teams find themselves outgrowing Node.js and TypeScript on the backend when building systems of consequence. In particular, TypeScript helps at dev and build time, but of course it's just JavaScript at runtime with all of its pitfalls and potential issues due to the lack of a strong static type system.

If we look at the lifecycle of a codebase from dev-to-build-to-runtime, the problem becomes clear:

Lifecycle	TypeScript	C#
Dev	Types present and help developers determine correctness	Ditto!
Build	Types inspected across the codebase to ensure correctness of the application as the codebase is transformed into JavaScript which is what is actually shipped out.	C# is compiled into an intermediate language (IL) or native binary (via ahead of time compilation (AOT)) that runs on the .NET runtime.
Runtime	Types no longer present 😱 to enforce correctness of data so another mechanism is needed (schema validation) to ensure correctness and guard against invalid data from sneaking in. Zod, Valibot, or other schema validators are necessary.	Type metadata still present and enforced at runtime 💪! Runtime errors happen when there's a type mismatch at a boundary (typically serialization). The difference is that this is "free" as a JSON payload is transformed into managed C# objects.

Should you use C# for web front-ends?

We're only considering backends here; I do not think that .NET-based front-ends (e.g. Blazor) are competitive in all use cases.

But even at dev time, TypeScript's fluid and flexible type system at scale can be problematic and affect productivity especially when teams are not disciplined about explicit types. This can often yield cases where it is necessary to trace a field down several layers of types before finding the actual type. It also affects how well refactoring works across large codebases where developers have heavily used utility types like Pick and Omit to create large, dependent type chains.

Once teams start to encounter these frictions points at scale, they will often consider alternatives such as:

• Go
• Java
• Kotlin
• Scala
• Rust (?!?)

But many teams end up overlooking .NET and C#. Some teams may even have an inherent bias against C# because of its "Micro$oft" roots while fully embracing other Microsoft products like GitHub, TypeScript, and VS Code! (the irony)

In fact, C# is possibly the most natural choice for teams that are already adept at TypeScript because the languages share so much in common. Both TypeScript and C# are designed by Anders Hejlsberg of Microsoft. Because of this lineage and the long influence of C# on JavaScript and vice versa, C# and TypeScript have actually been converging over the last decade. If you're building large backends in TypeScript, chances are you are already using Nest.js...which looks an awful lot like .NET controller web APIs.

One of the reasons is that many developers, engineering managers, and CTOs may have come across C# and .NET Framework at some point in their career, but haven't looked at it since Microsoft pivoted to the open source .NET Core initiative (that yielded the open source, numbered .NET versions). Today's .NET is very different from .NET Framework of the 2000's. Let's start with a simple distinction:

• .NET Framework - Win32 bindings, legacy, and primarily maintained and supported for legacy enterprise use cases. You'll see this as .NET Framework 4.8.x
• .NET 5, .NET 6, .NET 7, .NET 8, .NET 9, etc - Modern, cross-platform .NET that runs on Linux, macOS, and Windows with x64 or Arm64 targets.
• .NET Core, .NET Standard - Designator of the bifurcation point between legacy and modern (long legacy of terrible nomenclature from Microsoft...)

What's in a name?

I often wonder if Microsoft would have been better off just calling it dot instead when they made the transition from .NET Framework...

✨ Take me to the highlights! ✨

If you just want the highlights, I recommend:

• Collections
• Classes
• Express vs Minimal API
• Nest.js vs Controller API
• Databases and ORMs

The Problem with TS

Don't get me wrong: for working with JavaScript, TypeScript is an absolute necessity once a project grows beyond a toy and every JavaScript project I work on is TypeScript (e.g. CodeRev.app). Every JavaScript project that isn't a library should be written in TypeScript.

The fundamental problem with TypeScript-based backends and APIs is that this API:

export class CreateCatDto {
  name: string;
  breed: string;
}

export class CreateDogDto {
  name: string;
  breed: string;
  bark: "loud" | "medium" | "quiet"
}

// Nest.js controller, but could also be Express.js handler
@Post()
async create(@Body() createCatDto: CreateCatDto) {
  // Add a cat to the database
}

Will happily accept this payload at runtime:

{
  name: "Poochie",
  breed: "Chihuahua",
  bark: 42
}

It will also accept this:

{
  name: "Poochie",
  breed: "Chihuahua",
  bark: "eW91IHByb2JhYmx5IGRvbid0IHdhbnQgdGhpcyB1bnNhZmUgYW5kIG1hbGljaW91cyBwYXlsb2FkIGluIHlvdXIgZGF0YWJhc2UuLi4="
  // 👆 Would be passed through to a document-oriented database!
}

And now the createCatDto is carrying an extra bark property of dubious content because JavaScript doesn't care! TypeScript's type safety means nothing at runtime.

That's because the type information no longer exists at runtime and there's no enforcement of type which requires adding schema validators like Zod or Valibot to actually check the incoming payload conforms to some shape. Of course we can add validations and schemas to prevent a CreateCatDTO from accepting a CreateDogDTO at runtime, but this is extra work. (In fact, you might be here exactly because you're fed up with this extra work to ensure the correctness and safety of your backend application!)

If you are already using Nest.js...

Then you will probably like .NET controller web APIs which is conceptually similar with controllers, routes, middleware, and more.

The Cal.com codebase is a good example of this in action. Here's an example of a controller action "done right":

// organizations-schedules.controller.ts
@Roles("ORG_ADMIN")
@PlatformPlan("ESSENTIALS")
@UseGuards(IsUserInOrg)
@Patch("/users/:userId/schedules/:scheduleId")
@DocsTags("Orgs / Users / Schedules")
@ApiOperation({ summary: "Update a schedule" })
async updateUserSchedule(
  @Param("userId", ParseIntPipe) userId: number, // 👈 Note the validation for int
  @Param("scheduleId", ParseIntPipe) scheduleId: number, // 👈 Note the validation for int
  @Body() bodySchedule: UpdateScheduleInput_2024_06_11
): Promise<UpdateScheduleOutput_2024_06_11> {
  const updatedSchedule = await this.schedulesService.updateUserSchedule(
    userId, scheduleId, bodySchedule
  );

  return {
    status: SUCCESS_STATUS,
    data: updatedSchedule,
  };
}

And an example of a model using class-validator for payload validation:

import { CreateOrganizationAttributeOptionInput } from "@/modules/organizations/attributes/options/inputs/create-organization-attribute-option.input";
import { ApiProperty, ApiPropertyOptional } from "@nestjs/swagger";
import { AttributeType } from "@prisma/client";
import { Type } from "class-transformer";
import {
  IsArray,
  IsBoolean,
  IsEnum,
  IsNotEmpty,
  IsOptional,
  IsString,
  ValidateNested,
} from "class-validator";

export class CreateOrganizationAttributeInput {
  @IsString() // JS needs to be explicitly told this is a string
  @IsNotEmpty()
  @ApiProperty()
  readonly name!: string; // 👈 Even though we already have this type annotation

  @IsString()
  @IsNotEmpty()
  @ApiProperty()
  readonly slug!: string;

  @IsEnum(AttributeType)
  @IsNotEmpty()
  @ApiProperty({ enum: AttributeType })
  readonly type!: AttributeType;

  @IsArray()
  @ValidateNested({ each: true })
  @Type(() => CreateOrganizationAttributeOptionInput) // JS needs to be explicitly informed about this type
  @ApiProperty({ type: [CreateOrganizationAttributeOptionInput] })
  readonly options!: CreateOrganizationAttributeOptionInput[];

  @IsBoolean()
  @IsOptional()
  @ApiPropertyOptional()
  readonly enabled?: boolean;
}

Without the basic facilities for effective runtime type checking, JavaScript at runtime is entirely dependent on extra work to inform the runtime about type metadata. Another example of this is Cal.com's event-types.ts zod schemas which shows a different pattern using Zod schemas (this being their V1 API versus the previous V2 API example above):

// From the Cal.com codebase
const schemaEventTypeEditParams = z
  .object({
    title: z.string().optional(),
    slug: z
      .string()
      .transform((s) => slugify(s))
      .optional(),
    length: z.number().int().optional(),
    seatsPerTimeSlot: z.number().optional(),
    seatsShowAttendees: z.boolean().optional(),
    seatsShowAvailabilityCount: z.boolean().optional(),
    bookingFields: eventTypeBookingFields.optional(),
    scheduleId: z.number().optional(),
  })
  .strict();

// What this might look like in C#
record EventTypeEditParams(
  string? Title,
  string? RawSlug,
  int? Length,
  int? SeatsPerTimeSlot,
  bool? SeatsShowAttendees,
  bool? SeatsShowAvailabilityCount,
  EventTypeBookingFields[]? BookingFields,
  int? ScheduleId
) {
  public string? Slug => Slugify(this.RawSlug)
}

// The type itself enforces the data types
// It is not possible to assign `"hello"` to `.Length`

Effortless type safety feels like it's table stakes for what you want from a backend runtime. Certainly it doesn't matter for small, hobby projects, but beyond those use cases, JavaScript's "simplicity" is quickly subsumed by the complexity of mimicking rich runtime types.

Why C#

C# solves some of these problems that teams run into at scale with systems of consequence because it retains type metadata at runtime and enforces basic type alignment, eliminating some grunt work on the backend. But it's also remarkably similar to TypeScript in syntax while providing higher performance and throughput. It feels like the perfect alternative for teams burned out from TypeScript and Node.js.

"Didn't you see? The TypeScript team at Microsoft used Go for the compiler!"

Indeed, C# is not the right choice for every project and in the case of the TypeScript compiler, Go was the right tool for the right job and matched the existing coding style better than C# and .NET. While you can write CLI's in C# and .NET and you can write serverless functions in C#, neither use case are where it shines. But C# is a great choice for teams building serious backend APIs.

Sam Cox of Tracebit has a great writeup on why they chose C# for the backend and outlines many of the motivations I have found to be compelling for using C# (not despite, but because of my experience with both C# and TypeScript on the backend):

Professionally, I’ve worked in large and critical codebases in Python and TypeScript. I’ve seen the value they can bring (ecosystem, hiring, Engineer ramp times) but I’ve also experienced what felt like real drawbacks (dynamic typing, packaging & dependencies, maintaining large codebases).

If I had to define the key criterion I was looking for in a programming language, it would be Productivity. In general, I think a stack that lets you feel like you’re spending your time on what matters is the essence of a good developer experience.

It was important to me that the language was statically-typed. Having worked with both statically and dynamically-typed languages before, I believe that any ‘overhead’ that static typing introduces when writing code is repaid many times over.

Here are a few more reasons to consider C# and .NET for your next backend application:

It's Easy to Learn

Throughout this guide, you'll notice just how similar TypeScript and C# are because of their shared designer. This makes C# one of the easier languages to learn compared to Go or Scala if you already now TypeScript. I don't think it's a coincidence that C# is used in game engines like Unity and Godot; it offers the right amount of structure vs DX/ergonomics vs ease of adoption to make it broadly appealing.

Strongly Typed Yet Flexible

It turns out that -- if you are doing anything of consequence -- you actually probably want to have strong, static types on your backend at runtime because this will help reduce mistakes and errors while also reducing verbosity in the overall codebase required in JS to ensure some level of type safety (e.g. Zod schemas).

But .NET has a few interesting tricks up its sleeve when it comes to types such as:

• Type inference
• Tuples
• Named tuples
• Anonymous types
• Type extensions
• Union types (via 3rd party extensions for now...)

Read more in the section on classes and types

Balance of Runtime Performance + Great DX

• C# is not as fast and memory efficient as Rust
• Because it's JIT, it compiles to larger binaries than Go
• It also means it starts up slower than JavaScript

But for all of that, .NET still performs extremely well across the board in a variety of applications and with ahead-of-time compilation (AOT), it is possible to achieve relatively fast startup times.

For backends, .NET offers high performance roughly in the same ballpark with Java and Go with an excellent DX because of the large standard library and first party packages from Microsoft ("batteries included") as well as hot-reload (see CLI and Tooling).

Teams often think about performance incorrectly...

Thinking about performance in terms of response time is only one facet; the other is throughput or, for a given spend on infrastructure, how many requests can you handle? C# and .NET generally wins handily in both (with the exception of cold starts) because of its multithreaded runtime

Tooling

.NET offers great tooling support in VS Code (free) and Rider (free personal, paid commercial) and is supported on Linux, macOS, and of course Windows. It supports targeting both x64 and Arm64 architectures.

The dotnet cli is easy to use as well.

Read more in the next section on getting started.

Documentation

Microsoft's excellent documentation is cohesive and thorough, making it easy to read and understand. In Node, you might read documentation across a variety of third party sites for different packages (Zod, Prisma, Nest.js) which have different levels of thoroughness, different navigation, different writing styles, and so on. .NET's large standard library means that much of it also has consistent first party documentation.

Perhaps more importantly, it means that you'll likely get better results with LLMs because of the high quality documentation.

Consistency

One key issue with the rich ecosystem of Node.js is that no two projects look the same from a package and tooling perspective, even for very common workloads and foundational needs. On the other hand, .NET's rich first-party libraries means that more often than not, projects are going to look very consistent in terms of the libraries and tooling. Let's face it: CRUD APIs are CRUD APIs; having consistency can be a big boon to productivity by allowing developers to pick up common libraries quickly.

Consistency and productivity is a big benefit of having the platform ship a broad and deep baseline set of feature/functions while the OSS ecosystem around .NET is dominated by broadly used and battle-tested libraries. This aspect of C# and the .NET platform are very highly underrated for teams that are focused on shipping!

Security

JavaScript's lack of a deep and broad standard library means that a lot of functionality depends on third parties. On the other hand, .NET's large first party ecosystem means that if there are security issues, they will get patched. With third party libraries in either .NET or Node.js, this can be a roll of the dice!

Historically, NPM has also been a known vector for supply chain attacks and JavaScript's nature as a dynamic language means that it is susceptible to types of attacks that are not possible in a .NET runtime.

Check out GitHub's 2020 State of the Octoverse Security report for more insights.

Mature but Constantly Evolving

C# and .NET are constantly evolving and the language consistently improves DX with each release, reducing the verbosity of the language. It's akin to the relationship between Kotlin and Java on the JVM, except C# has continued to rapidly evolve as a language.

Check out some of the latest releases:

• C# 13
• C# 12
• C# 11
• C# 10

What should be clear is how much of a focus the .NET team has in terms of improving ergonomics and DX with each C# version.

Common Myths

TIP

There are a lot of myths about C# because many folks might have encountered it in the .NET Framework days and never looked at it afterwards. Let's look at some of the most common ones.

Myth: C# Requires Visual Studio and Expensive Licenses

This is not the case; in fact C# works great from VS Code! It's completely free to use and you can use the C# extensions without a license when using the C# extension for VS Code.

Myth: .NET is Windows Only

This was true of .NET Framework which had bindings to the Win32 dlls. Since .NET Core, Microsoft has been making .NET truly cross platform and any of the "numbered versions" like .NET 6, 7, 8, 9 can all run on Linux, macOS, and Windows.

In fact, at one startup, we developed .NET on M1 MacBook Pros, built our containers in GitHub Actions, and deployed those in AWS on Arm64 t4g instances.

.NET is a great cross-platform backend.

Myth: .NET is a Legacy Platform

This is true of .NET Framework, but the numbered .NET versions (.NET 6, 7, 8, 9) are modern and the underlying platform evolves extremely fast and is perhaps more akin to Kotlin in that sense.

Myth: .NET is Hard to Learn

That's why this doc exists 😅 In fact, .NET's congruence with TypeScript means that developers that know JavaScript and TypeScript should be able to easily transition to C#. In fact, I would argue that C# is probably easier to learn because of the more strict type system.

Myth: .NET Web APIs are Complicated

If you are building an enterprise system, you are probably going to choose Nest.js. At that point, you are writing controller classes, using decorators, and utilizing dependency injection -- exactly the same as if you were using .NET Web APIs. Except in Node.js, you don't get the benefits of the type system at runtime and throughput of a multi-threaded runtime.

Myth: .NET's Performance Over Node.js Doesn't Matter

Performance is probably the wrong way to think about the delta between something like .NET versus Node.js; it's probably better to think of it as throughput. Given a certain amount of hardware spend, .NET will yield a higher request throughput compared to Node.js solutions. This equates to less infra spend as you scale up and this can be significant.

Myth: C# is for Enterprises

While it's true that C# is an excellent choice in the enterprise, it's also used in game engines like Unity and Godot. This myth may have roots in the reality that early on in .NET Framework's life, the only tooling available was enterprise licenses for Visual Studio. The modern .NET tooling in VS Code and Rider makes working with C# free and a really great experience whether you're an indie developer, small startup, mid-size business, or enterprise scale!

Ready to Dive In?

If any of this rang true to you, let's dive in and see why C# and .NET might be right for your team!