Fastify, one of the most popular web frameworks for Node, just announced a major new release, version 5.0.
We were initially very excited to see if any cool new features or improvements were added, in case there was something we could take inspiration from in Encore.ts.
Unfortunately, after taking it for a spin, we felt it was quite underwhelming: the new release is extremely light on new features, but extremely heavy on breaking changes.
Let's take a look at what's new, and whether it's worth upgrading.
Fastify v5 comes with lots of breaking changes
Don't believe me? Check out the migration guide. It clocks in at almost 2000 words! It lists no less than 20 breaking changes. That's a lot of breakage to consider.
As far as we can tell, most of the breaking changes amount to "cleaning things up" in the code base.
It removes a lot of convenient shorthands for defining JSON Schemas, making schema validation more verbose and less readable.
It also removes a lot of deprecated functionality. Sometimes this is a good thing, if the functionality carries a significant maintenance burden. But in this case, a lot of the removed functionality is essentially a one-liner.
For example, reply.getResponseTime() is removed in favor of reply.elapsedTime. Maybe elapsedTime is a better name, but the cost of maintaining getResponseTime, with its one-line implementation, is zero.
Still, it's easy enough to grep a codebase for getResponseTime and fix it. Other breaking changes are more difficult.
Consider for example fastify.hasRoute() still exists, with exactly the same signature, but the behavior has changed in a subtle way. It used to support passing in a request path to evaluate whether an incoming request matches a route or not.
In Fastify v5 the method instead only supports matching the exact string given when registering the route. Migrating over to this is much more difficult, as it's not always clear whether you need to change your code or not.
Fastify v5 has very few, minor improvements
At this point I hear you say, "surely all the breaking changes were done for good reason!". Unfortunately the documentation is surprisingly sparse when it comes to the benefits of upgrading to v5.
The only new feature listed is support for the Diagnostic Channel API.
Is that worth upgrading for? For most people, probably not.
The other improvement is a modest performance boost.
From our benchmarking, Fastify v5 is about 5-10% faster than v4, depending on configuration settings and whether you use schema validation. That's nothing to sneeze at! But it's not a game changer either.
Is upgrading worth the hassle? Probably not.
It's disappointing to see a major new release of a popular framework that is so light on new features, and so heavy on breaking changes.
Particularly when the breaking changes are seemingly done to clean up internals rather than improving the user experience.
Our belief is that migrating from Fastify v4 to Fastify v5 is probably not worth it for most people.
If you care about performance or reliability, you should probably check out Encore.ts instead of spending time migrating to Fastify v5.
It's about twice as fast, and provides a lot of other benefits as well, like TypeScript-native schema validation, automatic tracing (no need to manually instrument your code with the Diagnostic Channel API!), and more.
What's more, Encore.ts takes backwards compatibility extremely seriously, and follows the Linux Kernel model when it comes to breaking changes: don't.
And Encore.ts is of course also open source.
Top comments (7)
Note this article is pretty biased
Still don't understand why they didn't take this opportunity to comply with web standards (e.g.
Request/ResponseAPIs).Web standards are frontend based, Fastify is a server framework.
I don't really see the point here. Web standards are web standards, regardless of where they're implemented. For instance, the fetch API has been adapted for Node.js to provide cross-runtime compatibility. Hono.js is also a server framework, but relies on standard Request and Response APIs 🤷
The web standards are meant for the web, not for servers. Also, it uses most standards (headers, url, ect.).
A 10% performance boost is really worth upgrading, since most of the changes were already deprecated in v4
Yeah I get they want to promote their product (which I can't really speak about), but cleaning up code and making it faster in the process is crucial, especially dropping support for Node <20.
The focus on performance has been notable on Node's side lately, so it's likely to boost Fastify's too. Plus it will be easier to develop new features. I would be slightly disappointed if v5.1.0 and beyond don't bring some interesting features, but not about this change.
If you were prepared and/or only did basic stuff (simple API with type validation), it was 0 effort and a free 10% performance improvement.