If you’re a web developer, please vote in Paul Irish’s poll on browser support. The larger the population that votes, the more we can trust the answers, and the data is critical to making sense of how we collectively think about browser support. Go now. It won’t take long, I promise.

And so we’re at an impasse.

At the last TC39 meeting, after spending what felt like an eternity on the important topic of UTF-16 encoding, the last day hastily ended with two topics that simply could not be any more urgent: can we get something done about data mutation observation — the underpinnings for every data binding system worth a damn — and classes.

As you might imagine, classes got at least one cold shoulder, but this time we were able to suss out the roots of the objection: the latest majority-agreed proposal (“Maximally Minimal Classes“) isn’t “high water mark” classes and, since so much has been given up to get something moved down the field, classes are no longer worth having. In this post I’ll outline a bit of the debate so far, current state, and hopefully convince you that the anti-classicist’s arguments are weak.

But first: why classes at all?

There’s an ambient argument in the JS world that “we don’t need no stinking classes”. That all we must do to save the language is teach people the zen of function objects and closures and all will be well with the world…and if we lose some people, so what? I don’t want to tar everyone who sympathizes with the beginning of that sentiment with the obvious negatives of the last bit, but that’s the overall gist I get. In this view, classes are an unnecessary and lead people to consider JS code the “wrong” way. I’ve written before about how in ES6 class means function, but that doesn’t mollify the discontent.

But then why any language feature at all? Why isn’t assembler good enough for everyone? Or C? Or C++?

Turns out the answer is “human frailty”. Or put a different way, the process of cognition depends on very limited amounts of short-term stack space in our wetware, and computing languages are about making the computer hospitable for the human, not about telling the system what to do. Our tradeoffs in languages would look much different if we could all easily recall 20 or 30 things at a time instead of 5-10. Languages are tools for freeing creative people from registers and stacks and heaps and memory management and all the rest; all while trying to keep the creative process grounded in the reality that it’s memory words in a Von Neumann architecture; without that grounding we’d end up too disconnected from the system to deliver anything practical.

Abstractions, high-level semantics, types…they’re all pivots, ways of considering sets of things while working with an individual example of that thing mentally. You don’t consider every case all at once and don’t remember everything about the system at one time; instead you focus on the arguments to the current function which creates it’s own scope. All of these mechanisms create ways of limiting what you need to think about. They ease the burden of associating some things with other things by allowing you to consider smaller cases and allowing you to create or break limitations as they become variously helpful and harmful to getting something done. And we have lots of words to describe the things we find painful about dealing with the overhead of remembering; of a lack of directness and locality: “spaghetti code”, “unmaintainable”, “badly factored”, “tightly coupled”, etc. Where there’s a new fad in programming, odds are high that it is being touted as a solution to some aggregate behavior which causes a reasonable local decision to become globally sub-optimal. This is what language features are for and why, when you see yourself using the same patterns over and over again, something is deeply wrong.

I’ll say this again (and imagine me saying it sloooooowly): languages evolve in response to the pain caused by complexity. The corollary is that language which don’t evolve to remove complexity, or which add it in day-to-day usage cause pain. Pain is almost always bad.

Complexity takes many forms, but the process of disambiguation when looking at a syntactic form that relies on a pattern is one of the easiest to take off the table. You’ll see this in my work in lots of areas: Web Components make the relationship between markup and eventual behavior less ambiguous when you read the HTML, Shadow DOM creates an isolation boundary for CSS and HTML that make it easier to consider a “thing” in isolation when building and using them, classes in JS make it easier to read some bit of code and not be tripped up by the many meanings of the word function, and many of the features I’ve advocated for in CSS (mixins, hierarchy, variables) are factoring mechanisms that make it simpler to pivot between repeated stuff and the ad-hoc visual semantics of the document that mean so much to the design language.

Not only should we write off the luddites, we should consider every step towards lower complexity good and steps towards complexity and slower understanding to be bad and actively harmful. Tortured claims about how “you aren’t going to need those things” need to be met with evidence; in this case, what do the pervasive libraries in the ecosystem do? Yeah, those are complexity to be taken off the table. The agenda is clear.

But back to classes and JS.

We’ve been debating various forms of classes in JS since the late ’90s. That’s right; we’ve been circling this drain for more than a decade. In that time, nearly every permutation of class system that I’m familiar with has been debated as the rightful heir to the reserved class keyword. From a parallel inheritance structure (non-prototypal) to “just prototypes like we have them today”, from “classes as constructor body” to “classes as object literals with constructors”, from frozen to gooey and unbaked, from classes as nominal types to totally untyped, we’ve seen it all. In all of this, the only constant has been failure. TC39 has continued, for more than a decade, to fail to field any class syntax (yes, I have my slice of blame-cake right here). In that time, JS has gone from being important to being critical to the construction of large systems; both on the client and the server. As this has happened, we’ve seen function pressed into service not only as lambda and method, but module, class, and pretty much every other form of ad-hoc composition system that’s needed at any point. We’re overdue for adding language-level support for all of these things. The committee has shown great ability to extend and expand upon idioms already in the language, taking syntax as a starting point and using it create new room for reducing pain (the spread operator and destructuring are great examples). Yet still many on the committee, notably Luke Hoban of Microsoft and Waldemar Horwat of Google, worry that a form of class in ES6 that doesn’t make coffee for you while also shooting rainbows out of it’s behind won’t be worth the syntactic space it takes up. I would have worried about this too back in ’08 when I first attended a TC39 meeting, but no longer. This is a group that is good at incremental change. Rainbow-shooting technology takes a while to build; meanwhile, wouldn’t it be great if we could get a cup of joe around here? We’ve waited long enough to get something meaningful to build on and with. I want a hell of a lot more than what the Maximally Minimal Classes proposal adds, but they’re a starting point, and there will be more ES versions.

So that’s Argument #1: people will hate us if class doesn’t do enough.

True, perhaps, but they will hate us anyway. Add classes and Mikeal Rogers personally organizes the pitchfork and torch distribution points and leads the march. Don’t add classes and the world at large thinks of JS as a joke and a toy — where’d that giant library/transpiler get to, anyway? You need something after all. Add them and Mikeal can keep not using classes to his heart’s content, and if we add classes in the style I prefer (and which Max/Min Classes embody), he can use them as old-skool functions and never care how they were declared; all the while making JS more productive for folks who don’t want to buy into a whole lifestyle to get a job done.

What’s Argument #2? That instances created from classes aren’t frozen. That is to say, it’s still possible to add new properties to them later. You know, like every other JS object.

If you’re thinking “yes, but isn’t there Object.freeze() already?”, you’re not alone. What you’re witnessing here is a debate about what’s considered good, not what’s possible. You can manually freeze/seal objects to your heart’s content, but what the (tiny) minority that is strenuously making argument #2 is demanding is that nobody be allowed to use the word class to create a constructor function body unless the resulting object is automatically frozen. They are willing to hold up the entire train until this preference is mollifed, and are unhappy with any short syntax which does not bless their preferred style.

In fairness, I also would be grumpy to see things go a direction I don’t prefer, but consider the following syntaxes that could be deployed in a future version to create frozen classes, none of which are precluded by the Max/Min proposal:

// "const" causes instances to have their properties 
// frozen on exiting the constructor
const class Point {
  constructor() { ... }
}
 
// Declaring a const properties member causes the
// constructor to check for those only, freezing exiting
// the constructor
class Point {
  constructor() { ... }
  const properties {
     // Object literal
     ...
  }
}

And I’m sure you can think of a couple of others. The essential distinguishing feature, and the thing that is holding the entire train up, is that the word class without any embellishment whatsoever doesn’t work this way. Yes, I know it sounds crazy, but that’s where we’re at. You can help, though. Erik Arvidsson has implemented Max/Min classes in Traceur and you can try it out in the live repl. If you like/hate it, won’t you please let the es-discuss mailing list know? Or just post in the comments here for posterity. We know that these classes need many more features, but your thoughts about this syntax as a starting point couldn’t be more timely.

Some stuff is too good to leave in the shadows. On my Bedrock post, James Hatfield writes in with a chilling point, but one which I’ve been making for a long while:

”every year we’re throwing more and more JS on top of the web”

The way things are going in my world, we are looking at replacing the web with JS, not simply layering. At a certain point you look at it all and say “why bother”. Browsers render the DOM not markup. They parse markup. Just cut out the middle man and send out DOM – in the form of JS constructs.

The second part is to stop generating this markup which then must be parsed on a server at the other end of a slow and high latency prone communication channel. Instead send small compact instructions to the browser/client that tells it how to render and when to render. Later you send over the data, when it’s needed…

This is a clear distillation of what scares me about the road we’re headed down because for each layer you throw out and decide to re-build in JS, you end up only doing what you must, and that’s often a deadline-driven must. Accessibility went to hell? Latency isn’t great? Doesn’t work at all without script? Can’t be searched? When you use built-ins, those things are more-or-less taken care of. When we make them optional by seizing the reigns with script, not only do we wind up playing them off against each other (which matters more, a11y or latency?) we often find that developers ignore the bits that aren’t flashy. Think a11y on the web isn’t great now? Just wait ’till it’s all JS driven.

It doesn’t have to be this way. When we get Model Driven Views into the browser we’ll have the powerful “just send data and template it on the client side” system everyone’s looking for but without threatening the searchability, a11y, and fallback behaviors that make the web so great. And this is indicative of a particularly strong property of markup: it’s about relationships. “This thing references that thing over there and does something with it” is hard for a search engine to tease out if it’s hidden in code, but if you put it in markup, well, you’ve got a future web that continues to be great for users, the current crop of developers, and whoever builds and uses systems constructed on top of it all later. That last group, BTW, is you if you use a search engine.

But it wasn’t all clarity and light in the comments. Austin Cheney commented on the last post to say:

This article seems to misunderstand the intention of these technologies. HTML is a data structure and nothing more. JavaScript is an interpreted language whose interpreter is supplied with many of the most common HTML parsers. That is as deep as that relationship goes and has little or nothing to do with DOM.

…It would be safe to say that DOM was created because of JavaScript, but standard DOM has little or nothing to do with JavaScript explicitly. Since the release of standard DOM it can be said that DOM is the primary means by which XML/HTML is parsed suggesting an intention to serve as a parse model more than a JavaScript helper.

Types in DOM have little or nothing to do with types in JavaScript. There is absolutely no relationship here and there shouldn’t be…You cannot claim to understand the design intentions around DOM without experience working on either parsers or schema language design, but its operation and design have little or nothing to do with JavaScript. JavaScript is just an interconnecting technology like Java and this is specifically addressed in the specification in Appendix G and H respectively.

And, after I tried to make the case that noting how it is today is no replacement for a vision for how it should be, Austin responds:

The problem with today’s web is that it is so focused on empowering the people that it is forgetting the technology along the way. One school of thought suggests the people would be better empowered if their world were less abstract, cost radically less to build and maintain, and is generally more expressive. One way to achieve such objectives is alter where costs exist in the current software life cycle of the web. If, for instance, the majority of costs were moved from maintenance to initial build then it could be argued that more time is spent being creative instead of maintaining.

I have found that when working in HTML alone that I save incredible amounts of time when I develop only in XHTML 1.1, because the browser tells you where your errors are. … Challenges are removed and costs are reduced by pushing the largest cost challenges to the front of development.

… The typical wisdom is that people need to be empowered. If you would not think this way in your strategic software planning then why would it make sense to think this way about strategic application of web technologies? …

This might all sound very rational on one level, but a few points need to be made:

• If we’re not building technology for people, WTF are we doing with these CPU cycles exactly?
• Speed of iteration is a key enabler of progress in any technology stack I’ve ever worked in
• Strictness fails in the wild

I think Austin’s point about moving costs from maintenance to build is supposed to suggest that if we were only more strict about things, we’d have less expensive maintenance of systems, but it’s not clear to me that this has anything to do with strictness. My observation from building systems is that this has a lot more to do with being able to build modular, isolated systems that compose well. Combine that with systems that let you iterate fast, and you can grow very large things that can evolve in response to user needs without turning into spaghetti quite so quickly. Yes, the web isn’t great for that today, but strictness is orthogonal. Nothing about Web Components demands strictness to make maintainability infinitely better.

And the last point isn’t news. Postel’s Law isn’t a plea about what you, dear software designer, should be doing, it’s an insightful clue into the economics of systems at scale. XML tried being strict and it didn’t work. Not even for RSS. Mark Pilgrim’s famously heroic attempts at building a reliable feed parser match the war stories I’ve heard out of the builders of every large RSS system I’ve ever talked to. It’s not that it’s a nice idea to be forgiving about what you accept, it’s that there’s no way around it if you want scale. What Austin has done is the classic bait-and-switch: he has rhetorically substituted what works in his organization (and works well!) for what’s good for the whole world, or even what’s plausible. I see this common logical error in many a standards adherent/advocate. They imagine some world in which it’s possible to be strict about what you accept. I think that world might be possible, but the population would need to be less than the size of a small city. Such a population would never have ever created any of the technology we have, and real-world laws would be how we’d adjudicate disputes. As soon as your systems and contracts deal with orders of magnitude more people, it pays to be reliable. You’ll win if you do and lose if you don’t. It’s inescapable. So lets banish this sort of small-town thinking to the mental backwaters where it belongs and get on with building things for everyone. After all, this is about people. Helping sentient beings achieve their goals in ways that are both plausible and effective.

If helping people is not what this is about, I want out.

Dave Herman jokingly accused me a couple of TC39 meetings ago of being an “advocate for JavaScript as we have it today”, and while he meant it in jest, I guess to an extent it’s true — I’m certainly not interested in solutions to problems I can’t observe in the wild. That tends to scope my thinking aggressively towards solutions that look like they’ll have good adoption characteristics. Fix things that are broken for real people in ways they can understand how to use.

This is why I get so exercised about WebIDL and the way it breaks the mental model of JS’s “it’s just extensible objects and callable functions”. It’s also why my discussions with folks at last year’s TPAC were so bleakly depressing. I’ve been meaning to write about TPAC ever since it happened, but the time and context never presented themselves. Now that I got some of my words out about layering in the platform, the time seems right.

Let me start by trying to present the argument I heard from multiple sources, most likely from (in my feeble memory) Anne van Kestern Jonas Sicking(?):

ECMAScript is not fully self-describing. Chapter 8 drives a hole right through the semantics, allowing host objects to whatever they want and more to the point, there’s no way in JS to describe e.g. list accessor semantics. You can’t subclass an Array in JS meaningfully. JS doesn’t follow it’s own rules, so why should we? DOM is just host objects and all of DOM, therefore, is Chapter 8 territory.

Brain asploded.

Consider the disconnect: they’re not saying “oh, it sure would be nice if our types played better with JS”, they’re saying “you and what army are gonna make us?” Remember, WebIDL isn’t just a shorthand for describing JavaScript classes, it’s an entirely parallel type hierarchy.

Many of the Chapter 8 properties and operations are still in the realm of magic from JS today, and we’re working to open more of them up over time by giving them API — in particular I’m hopeful about Allen Wirfs-Brock’s work on making array accessors something that we can treat as a protocol — but it’s magic that DOM is appealing to and even specifying itself in terms of. Put this in the back of your brain: DOM’s authors have declared that they can and will do magic.

Ok, that’s regrettable, but you can sort of understand where it comes from. Browsers are largely C/C++ enterprises and DOM started in most of the successful runtimes as an FFI call from JS to an underlying set of objects which are owned by C/C++. The truth of the document’s state was not owned by the JS heap, meaning every API you expose is a conversation with a C++ object, not a call into a fellow JS traveler, and this has profound implications. While we have one type for strings in JS, your C++ side might have bstr, cstring, wstring, std::string, and/or some variant of string16.

JS, likewise, has Number while C++ has char, short int, int, long int, float, double, long double, long long int…you get the idea. If you’ve got storage, C++ has about 12 names for it. Don’t even get me started on Array.

It’s natural, then, for DOM to just make up it’s own types so long as its raison d’être is to front for C++ and not to be a standard library for JS. Not because it’s malicious, but because that’s just what one does in C++. Can’t count on a particular platform/endianness/compiler/stdlib? Macro that baby into submission. WTF, indeed.

This is the same dynamic that gives rise to the tussle over constructable constructors. To recap, there is no way in JS to create a function which cannot have new on the left-hand-side. Yes, that might return something other than an instance of the function-object on the right-hand side. It might even throw an exception or do something entirely non-sensical, but because function is a JavaScript concept and because all JS classes are just functions, the idea of an unconstructable constructor is entirely alien. It’s not that you shouldn’t do it…the moment to have an opinion about that particular question never arises in JS. That’s not true if you’re using magic to front for a C/C++ object graph, though. You can have that moment of introspection, and you can choose to say “no, JS is wrong”. And they do, over and over.

What we’re witnessing here isn’t “right” or “wrong”-ness. It’s entirely conflicting world views that wind up in tension because from the perspective of some implementations and all spec authors, the world looks like this:

Not to go all Jeff Foxworthy on you, but if this looks reasonable to you, you might be a browser developer. In this worldview, JS is just a growth protruding from the side of an otherwise healthy platform. But that’s not how webdevs think of it. True or not, this is the mental model of someone scripting the browser:

The parser, DOM, and rendering system are browser-provided, but they’re just JS libraries in some sense. With <canvas>‘s 2D and 3D contexts, we’re even punching all the way up to the rendering stack with JS, and it gets ever-more awkward the more our implementations look like the first diagram and not the second.

To get from parser to DOM in the layered world, you have to describe your objects as JS objects. This is the disconnect. Today’s spec hackers don’t think of their task as the work of describing the imperative bits of the platform in the platform’s imperative language. Instead, their mental model (when it includes JS at all) pushes it to the side as a mere consumer in an ecosystem that it is not a coherent part of. No wonder they’re unwilling to deploy the magic they hold dear to help get to better platform layering; it’s just not something that would ever occur to them.

Luckily, at least on the implementation side, this is changing. Mozilla’s work on dom.js is but one of several projects looking to move the source of truth for the rendering system out of the C++ heap and into the JS heap. Practice is moving on. It’s time for us to get our ritual lined up with the new reality.

Which brings me too David Flanagan who last fall asked to read my manifesto on how the platform should be structured. Here it is, then:

The network is our bottleneck and markup is our lingua-franca. To deny these facts is to design for failure. Because the network is our bottleneck, there is incredible power in growing the platform to cover our common use cases. To the extent possible, we should attempt to grow the platform through markup first, since markup provides the most value to the largest set of people and provides a coherent way to expose APIs via DOM.

Markup begets JS objects via a parser. DOM, therefore, is merely the largest built-in JS library.

Any place where you cannot draw a line from browser-provided behavior from a tag to the JS API which describes it is magical. The job of Web API designers is first to introduce new power through markup and second to banish magic, replacing it with understanding. There may continue to be things which exist outside of our understanding, but that is a challenge to be met by cataloging and describing them in our language, not an excuse for why we cannot or should not.

The ground below our feet is moving and alignment throughout the platform, while not inevitable, is clearly desirable and absolutely doable in a portable and interoperable way. Time, then, to start making Chapter 8 excuses in the service of being more idiomatic and better layered. Not less and worse.