I was recently on Jason Swett’s podcast again. He’s a great interviewer and I always have fun with him.
By Twitter request we talked about… When would you not use Rails? It’s a great question.
For the entertaining version, listen to the podcast. For the just-the-facts extra-complete version, I’m writing this post.
When Is Rails the Wrong Choice?
I’ll start with a few simple, obvious times you wouldn’t use Rails, and then I’ll talk about some technically interesting times.
First, and most important, is team familiarity. If your team doesn’t already know Rails and isn’t especially interested in learning it then Rails is the wrong choice. This should be obvious, but it still deserves first billing.
Second, when you know some other framework fits better. I’ll talk more below about when that is. But sometimes you have a specific concern that trumps everything else. If you need to use a Java-language machine learning library and you don’t want to use JRuby for some reason, Rails isn’t your framework. If you’re writing a WordPress plugin, you’ll be doing it in PHP. Often there’s one specific compatibility concern that overrides everything else.
You can also think of it as: use it where Rails’ good points hold and its bad points don’t. So we’ll also talk about the good and bad points.
Separately: you’d normally only use Rails as an HTTP server, so some tasks just aren’t Rails-shaped.
When is Rails Too Much?
Some places not to use Rails can include:
Really Small Tasks that Won’t Grow: if a server does very little, Rails is often too much. Not going to touch a database? Then the DB setup isn’t helping you, is it? Just a tiny low-traffic intermediate server with no caching? A lot of Rails is more trouble than it’s worth.
Be careful with tasks that grow, though — making a tiny server scale up to do a lot more can be ugly. If you’re already serving HTTP pages to a human with a web browser, consider that you may have to add features to it later. Something like that is already fairly large from the word “go”.
Rails security really shines when you’re navigating the bewildering world of HTML and browser security. Small projects that mostly speak a structured format read by machines will get less from Rails. Securing something like an integer ID or a hash of strings is just easier than ensuring your HTML contains no script tags or anything exploitable.
Related to that is when you’re doing in-browser rendering and Rails is ‘just’ serving JSON. It’s a weird kind of in-between case. A lot of Rails security and convenience functions no longer help you, but you’re still doing things where internal libraries (ActiveRecord, ActiveJob, ActionMailer) can be highly useful. But if you’re never rendering HTML on the server and you’re very sure you never will, Rails will probably help you less.
When Is Rails Not Enough?
Rails is also designed for a small team and a medium-sized codebase. A huge team (lots of programmers) or a huge codebase (lots of controllers, models and/or lines of code) will tend to drag down the standard Rails-app structure.
Ruby allows for a lot of non-local effects. Whether that’s monkeypatching, writing to a database or creating new types at runtime, Ruby isn’t designed for a team of 200 programmers where you don’t trust some of them. There are too many ways for them to cause you trouble. You can use good tooling to scale Ruby to larger teams, but even that will tend to have exceptions and difficulties. That’s not really Ruby’s sweet spot.
In most cases you can cut up a large project into smaller projects. If one Rails app is too big, you can often separate it into multiple apps, or a thinner app with more back-end services, or an app and a separate microservice, or… One way or another there is usually a way to separate out smaller pieces. Ruby strongly encourages that, as do I.
There are also not-quite-Rails structures that can scale better. Avdi Grimm’s (now retired) Objects on Rails was an attempt in that direction, as is the Hexagonal architecture for Rails, which in turn has a lot in common with the older and more general N-tier architecture.
But at some point you might want to consider a different framework. Hanami is an obvious choice, designed to be less quick and nimble than Rails for getting a tiny app off the ground, but more scalable if you want to use the same code with a lot more contributors.
I’d still start out in Rails, personally. If you’re building something quickly to see if anybody cares, I know of no framework that comes close to its productivity. Wait to rewrite (in a more rigid framework) until you’re successful and you can afford the drag on your development speed.
The other worry here can be performance. If you’re rewriting a project that is already as large as the current Basecamp… then you’re actually fine for performance. Rails still scales great for them. But if you’re looking at something a hundred times larger (which by definition means B2C, not B2B) then you might have a situation where your server costs are substantially greater than your engineering payroll. In that case it can make sense to slow down your engineers to pay lower server costs. To check this, see what your EC2-or-equivalent costs are just for your application servers, which are what run Rails. And check your payroll just for web engineers, which is who writes in Rails. Normally the engineering payroll is much larger and you should stick with trading cheap machine time for expensive engineering time. But at some point the balance may tip and you should consider raising your engineering payroll to cut your server costs.
When Does Rails Have the Wrong Assumptions?
Before checking if Rails’ assumptions are right for you, we should see what those assumptions actually are.
Before you take my word for it, I recommend taking David Heinemeier Hansson’s word for it in the form of The Rails Doctrine. It’s a great document and it covers a lot of ground.
Indeed, if you want to better understand why Rails isn’t amazing for large, low-trust teams, you should read “Provide Sharp Knives” in the Rails Doctrine several times. A lot of Rails’ tradeoffs are entirely by design.
Rails also has some simpler assumptions: it assumes you’re writing an interactive app with server-rendered HTML. It assumes that security is vital (Rails trades a lot for security) but that you don’t want to build your own custom security system in most cases. And it assumes that you either have a small, excellent team doing prototyping work (“Provide Sharp Knives”) or that you have a possibly-mediocre team that needs powerful built-in guidelines (“The Menu is Omakase.”)
Rails also assumes you want high developer velocity at a cost of technical debt. In other words, it’s designed for building very quickly. That makes sense when technical execution is not your biggest risk. For instance: if you’re building a small startup, and you’re pretty sure you can build the site but people may not buy your product, you are dominated by market risk. That’s when Rails is perfect. You want to build very quickly. And even if you build perfectly, you’re probably going to have to throw away the result for nontechnical reasons, like “people don’t want to buy it.”
As part of “high dev velocity, technical debt is okay” Rails assumes things like, “you’ll want to use a lot of gems” and “dependencies that work are fine if they speed you up.”
Rails assumes you don’t mind scaling out application servers horizontally (by bringing more of them online.) It’s designed to scale well if you can do that. Rails assumes CPU is fairly cheap and it’s usually right about that. Relatedly, Rails assumes that the database is usually your most serious performance bottleneck, which is how web applications usually work.
Rails also assumes you’ll have some calculation or data transformation in your application. It assumes that it’s okay to use some CPU because you’ll be doing that anyway.
(When does that last assumption matter? Let’s talk about Evented Servers and see.)
What Isn’t Rails Good At?
While Rails is great at a lot of things, there’s one particular task that it’s not amazing for: shim servers.
By “shim servers” I mean servers that do very little calculation but integrate answers from a few other back-end services and relay the result. Imagine a server that queries two JSON services and combines the result with simple string-manipulation, for instance. It does very little calculation, but it juggles a lot of events.
And that’s the relevant word: “events.”
There is a specific kind of app architecture embodied by Node.js and its relatives called “Evented” programming. It can support many thousands, or even millions, of simultaneous connections with a tiny amount of server resources. It can be both high-throughput and low-latency. Its benchmark numbers are matchless… for the cases where it works.
If Evented is so much better, why don’t we use it for everything? Because it doesn’t work for everything. I emphasise calculation per-request because an Evented server will fall down and die if you try to make it do very much calculation per-request. Having one server handle a million connections is no good if each connection winds up using a hundred milliseconds of CPU time — that’s simply too many connections and the latency will be terrible.
In other words, Rails and Node.js are different tools for different projects. If you’re thinking, “I should either use Rails or Node for this” I would recommend looking deeper into your project (and/or your framework) until it’s obvious which one is the right answer. They do different things.
Look, I Just Scroll to the Bottom for the Summary and Criticise It On Reddit
Rails is the wrong choice if your team doesn’t want to use it or doesn’t know how.
Rails is the wrong choice in cases where a different framework is specifically better, or you have a specific library you need to be compatible with that isn’t Rails-friendly.
Rails might be the wrong choice if you’re not rendering HTML on the server, especially if your project is very small and/or doesn’t use a database.
Rails is the wrong choice is you’re not doing prototyping-flavoured work, preferably with a small, highly-competent team.
Rails is the wrong choice if your dev team or your app code is too big and you can’t subdivide the project.
Rails is the wrong choice if your project wants an Evented server like Node.js or EventMachine.
This article is the wrong choice if you’d rather listen to an entertaining podcast on the same topic.
If you’re wondering when Rails is the right choice, the Rails Doctrine is a great first step.