Scott Hanselman

JavaScript is Assembly Language for the Web: Part 2 - Madness or just Insanity?

July 19, '11 Comments [42] Posted in Javascript
Sponsored By

UPDATE: Check out the Podcast I did with Erik Meijer on Hanselminutes this week on this very topic: JavaScript is Assembly Language for the Web: Semantic Markup is Dead! Clean vs. Machine-coded HTML.

Some folks think that saying "JavaScript is Assembly Language for the Web" is a totally insane statement. So, I asked a few JavaScript gurus like Brendan Eich (the inventor of JavaScript) and Douglas Crockford (inventor of JSON) and Mike Shaver (Technical VP at Mozilla). Here is our private email thread, with their permission.

Mike Shaver:

I've heard the comparison before, and I think it's mostly right. It ignores the amount of effort put into JS's developer ergonomics, though, since assembly is not designed to have a humane syntax (especially modern assembly).

Brendan Eich:

I said "JS is the x86 of the web" a couple of years ago [likely at JSConf], but I can't claim it's original. [Nick Thompson said it on Hacker News this year as well.]

The point is JS is about as low as we can go. But it also has higher-level facilities.

Shaver's right, assembly without a great macro processor is not good for programmers or safety. JS is. So the analogy needs some qualification or it becomes silly.

The mix of high-level functional programming and memory safety, with low-level facilities such as typed arrays and the forthcoming ES.next extension of typed arrays, binary data, make for a more powerful programming language than assembly, and of course memory safety is the first differentiator.

Douglas Crockford:

I think it is a little closer to the mark to say that JavaScript is the VM of the web. We had always thought that Java's JVM would be the VM of the web, but it turns out that it's JavaScript.

JavaScript's parser does a more efficient job of providing code security than the JVM's bytecode verifier. JavaScript did a better job of keeping the write once, run everywhere promise, perhaps because it works at a higher level, avoiding low level edge cases. And then Turing takes care of the rest of it.

There are certainly a lot of people who refuse to consider the possibility that JavaScript got anything right. I used to be one of those guys. But now I continue to be amazed by the brilliance that is in there.

Brendan Eich, again:

Doug's point about source beating bytecode is good. My friend Prof. Michael Franz of UC Irvine long ago showed O(n^4) complexity (runaway compute cycles, denial of service) in the Java verifier. JS is strictly more portable and fast enough to lex/parse as minified source.

Source as "bytecode" also avoids the big stupid Java bytecode mistake: freezing a poorly designed lowered form of Java, then being unable to evolve the high-form source, i.e., the Java programming language for fear of breaking Java bytecode compatibility. This severely messed up the design of inner classes and then generics in Java -- and then Sun broke bytecode compat anyway!

From a YCombinator Thread a while back, Nick Thompson said:

My admittedly biased view: I spent two years of my life trying to make the JVM communicate gracefully with Javascript - there were plenty of us at Netscape who thought that bytecode was a better foundation for mobile code. But Sun made it very difficult, building their complete bloated software stack from scratch. They didn't want Java to cooperate with anything else, let alone make it embeddable into another piece of software. They wrote their string handling code in an interpreted language rather than taint themselves with C! As far as I can tell, Sun viewed Netscape - Java's only significant customer at the time - as a mere vector for their Windows replacement fantasies. Anybody who actually tried to use Java would just have to suffer.

Meanwhile Brendan was doing the work of ten engineers and three customer support people, and paying attention to things that mattered to web authors, like mixing JS code into HTML, instant loading, integration with the rest of the browser, and working with other browser vendors to make JS an open standard.

So now JS is the x86 assembler of the web - not as pretty as it might be, but it gets the job done (GWT is the most hilarious case in point). It would be a classic case of worse is better except that Java only looked better from the bottom up. Meanwhile JS turned out to be pretty awesome. Good luck trying to displace it.

The point is, of course, that no analogy is perfect. Of course JavaScript as a language doesn't look or act like ASM. But as an analogy, it holds up.

  • JavaScript is ubiquitous.
  • It's fast and getting faster.
  • Javascript is as low-level as a web programming language goes.
  • You can craft it manually or you can target it by compiling from another language.

This topic comes up on Hacker News often.

  • "The JavaScript we've got now is the assembly language of the client-side. We can't easily change it, but we have to start building better tools on top of it." - jonnycat

Have at it. I enjoy our thoughtful, measured and reasoned discussions, Dear Reader. You guys rock.

About Scott

Scott Hanselman is a former professor, former Chief Architect in finance, now speaker, consultant, father, diabetic, and Microsoft employee. He is a failed stand-up comic, a cornrower, and a book author.

facebook twitter subscribe
About   Newsletter
Sponsored By
Hosting By
Dedicated Windows Server Hosting by ORCS Web
Tuesday, July 19, 2011 6:51:57 PM UTC
Wow. Thanks for sharing that fragment of the conversation. It's very insightful.
Tuesday, July 19, 2011 6:57:26 PM UTC
My pleasure.
Tuesday, July 19, 2011 7:13:30 PM UTC
I didn't quite understand what you were talking about on your last post, but your podcast cleared it up for me. You may want to check out Coffeescript. It's a new language, similiar to Python/Ruby, that compiles down to javascript.
Tuesday, July 19, 2011 7:18:00 PM UTC
Mike - Yes, I'm all over Coffeescript. I have a post coming.

As for the last post, I think it was poorly written and that's my fault.
Tuesday, July 19, 2011 7:29:01 PM UTC
I enjoyed the exchange but was disappointed that they didn't address the most interesting part of the analogy. That JavaScript is a better VM than the JVM is interesting but not, to me, as interesting as the idea that JavaScript shouldn't be written by humans - it should be generated by compilers.
Tuesday, July 19, 2011 7:31:26 PM UTC
CoffeeScript is great. The fusion of Python and Ruby syntax is very pretty, easy on eyes and fingers. Its creator Jeremy Ashkenas was kind enough to let me co-present during the last JSConf.us (http://blip.tv/jsconf/jsconf2011-jeremy-ashkenas-5258082). One of the points that is a little computer-sciency, but important, about CoffeeScript and similar languages, is that the language's semantic level of expressiveness matches JS. The much more concise and convenient syntax may fool people into thinking there are "new semantics" in CoffeeScript, but this is not so in a formal sense.

Indeed mapping a language with novel semantics to JS requires more elaborate compilation and runtime support, and that can be a limiting factor. An example would be Continuation Passing Style conversion of functions using yield in a version of JS with generators (https://developer.mozilla.org/en/new_in_javascript_1.7, to be included in JS's next standard: http://wiki.ecmascript.org/doku.php?id=harmony:generators). Such a compiler has to turn generator functions into spaghetti coded state machines.

This is doable but it takes more work and it may cost more on the target JS VM. So JavaScript language evolution continues, in order to fill semantic gaps. And the smart choice of CoffeeScript to be "just syntax" means it will continue to work well, and it can be extended to map new JS semantics with Coffee-esque syntax, as appropriate.

/be
Tuesday, July 19, 2011 7:31:53 PM UTC
From my perspective, JavaScript CAN be generated by compilers, but not necessarily SHOULD be.
Tuesday, July 19, 2011 7:37:59 PM UTC
Kevin: I didn't "address" your claim that humans should not write JS because I don't believe it. Indeed shaver and I spoke about developer ergonomics and usability.

JS will remain hand-coded as well as generated for the foreseeable future. That's part of what helped it win too: people could view source (better done on github.com now, but still usable with http://jsbeautifier.org/).

/be
Tuesday, July 19, 2011 7:48:38 PM UTC
Scott,
JavaScript can definitely be generated by compilers, but since size and speed still matter, humans can generally do much better job. After all, writing JavaScript is not that much more difficult than writing c#. It is not like we need to create binary code.
Tuesday, July 19, 2011 7:56:57 PM UTC
It all seems like a natural evolution to me. We used to use rocks to hammer things in. Then we had hammers. Then we start building those hammers in factories with ther hammers. Now we have machines that build machines that eventually build craftsman auto-hammers.

It seems like when you have an application as large and complex as a facebook or google+ the problems with the JavaScript language become bigger issues. Using a higer level classical object oriented langauge with the benefits of type-safety are probably worth the trade off in increased complexity.

On the other hand, a medium sized site that can be maintained by hand full of devs probably is not.
ctrlShiftBryan
Tuesday, July 19, 2011 8:04:45 PM UTC
ctrlShiftBryan: as far as I know, Facebook does not translate from a classical OOP language to JS -- they hand-code. Google likewise does not use GWT much on its properties (please correct me if I'm wrong), although they do use the Closure compiler to translate from a lightly extended JS dialect to optimized/minified JS.

Believe it or not, classical OOP languages -- especially ones with not-very-expressive static type systems -- are not the only way to build large apps. It can be done in JS, but it's harder than it ought to be. This too is driving JS language evolution in Ecma TC39, where I spend a lot of my time (es-discuss at mozilla.org is the unofficial discussion list).

/be
Tuesday, July 19, 2011 8:19:20 PM UTC
Scott - one question that I wished you asked Eric on your podcast is 'what happened to Volta'? The door seems to be wide open to establish a Silverlight like higher level abstraction on top of Canvas and ECMAScript5.
Joe Wood
Tuesday, July 19, 2011 9:12:43 PM UTC
Javascript is like the INT21 of DOS, the escape pod out of a confined world, not a basis to build applications on.
There is still a lot missing, in the language or the browser runtime.
It is a cross-platform or rather cross-browser platform language for building website/applications, but that's as far the analogy goes.

Yes you can build a word processor in javascript, but at the end of the day, the products are not in the same class as the native.
Javascript and webapps enables the instant running without installing, and that's a very big plus, but is it enough. Where at a cross-road here, just as we were at the beginning of the 90s.

Marcel
MarcelDevG
Tuesday, July 19, 2011 9:59:47 PM UTC
Seems to me that nowadays we "write to jQuery". And while jQuery is actually a JavaScript framework, it seems like another language -- one that feels like it compiles to JavaScript.
Tuesday, July 19, 2011 10:58:16 PM UTC
Slapout - Totally. There are JavaScript programmers and then there are jQuery programmers! ;)
Wednesday, July 20, 2011 12:27:32 AM UTC
I actually came up in JavaScript and have started to study assembly. The analogy doesn't really work where it counts the most for me which is in the category of fine-grained control and not having anything denied or hidden from you.

I wish the ECMA specs would require that all built-in object/property labels be over-writable and exposed by for-in loops. That makes it a lot easier to kick the tires of new browsers and correct issues directly when certain vendors settle for "close enough" on their stalled version of the W3C DOM API spec for 10 years.

In spite of this, JavaScript is extremely malleable and it's that desire for freedom in my expanding repertoire that makes me want to know what's at the core of it all. The more I learn, the more I dislike languages that force a given paradigm on you. Static typing? No problem. Everything must be class-based? Uh... As an outside observer of the phenomenon who has occasionally been forced to work around some exceedingly inflexible Java, I can't help but notice that when you insist that everything is OOP, what tends to happen is that nothing is and you end up with is a whole lot of ungainly procedural code wearing class tutus. I'm not adverse to class-based approaches when they work well and when dealing data-intensive scenarios on the front end, they can certainly come in handy.

When the easily emulated classical OOP paradigm is less ideal, however, I am deeply grateful for the ability to throw object literals and functions with baked-in declaration contexts around like they're candy with a node-based markup language and the API for plugging into it being all the 'real-world' structure needed to bring it all together in a sensible fashion.
Wednesday, July 20, 2011 1:34:34 AM UTC
JS's lack of static typing is a limiting factor for raw performance.
Wednesday, July 20, 2011 4:53:01 AM UTC
Hi Scott, this is a great post. Can I have the permission to translate it into Chinese and post it on my blog (I will include the original link)?
Jia Mi
Wednesday, July 20, 2011 5:37:44 AM UTC
I would love to see a Microsoft supported C# to JS compiler, like Script#. Whenever I'm coding in C# and want to do similar things in JS, I always feel I need to take a step back. My mental model of what is possible is broken and I need to focus on the plumbing of JS, whereas I just would like to create a relatively simple object model. It would greatly enhance my productivity.
Wednesday, July 20, 2011 12:02:42 PM UTC
It's amazing that Javascript, a language that's almost venerable in age (especially when you consider that it's syntax cannot evolve quickly the way compiled languages can, and for practical purposes it's unchanged since it was invented), is only being exploited to anything close to its full potential in the last couple years. Can you imagine if you had suggested to someone that a game like Angry Birds could run in a web browser even three years ago?

jQuery and to a lesser extent other JS frameworks are responsible for this. jQuery isn't a different language. It's a framework. It's the .NET for C#. It's the rails for ruby. It *inspired* a generation of web developers to start caring about Javascript. Even using jQuery at the most basic level makes you learn about fundamental Javascript concepts. Using a callback? You just learned about anonymous functions and the concept of functions as first class objects. Using "each"? You just learned about "this" context. These are among the most difficult things for programmers of other languages to understand when learning Javascript.

While it is possible to get some use out of jQuery without understanding the details of Javascript syntax, it's not a different language, and it's a starting point for many people who get a taste of what they can do and decide they want to learn more. It's not watering down Javascript at all, it's exposing it to millions of people who would never have been exposed.
Wednesday, July 20, 2011 1:30:33 PM UTC
Using server gadgets to auto-generate client side code seems so last decade.

I believe we are at a turning point in web development. Using JQuery/Javascript, AJAX, CSS3, & HTML5 we can finally move functionality which belongs in the client to the client. I still see myself writing server side code, but not to compensate for my lack of client-side skills.

Finally, I agree with the previous posting: JQuery has given me the opportunity to appreciate Javascript.
bitFlinger
Wednesday, July 20, 2011 1:56:04 PM UTC
@Jamie "Can you imagine if you had suggested to someone that a game like Angry Birds could run in a web browser even three years ago?"

I did this mock up of Space Invaders around five years ago (and then quickly found it wasn't remotely original, lots of people had already done complete "coin-op conversions" in JS).

The thing that really gets me is the actual working JS emulation of enough x86 instructions to boot up Linux in the browser. Still half suspecting it to be a hoax!

Back to "the assembler of the web" - assembler is notoriously non-portable, whereas JS is surprisingly portable despite what people say.

The real analogy is with C, sometimes called the "portable assembler language".

It has long served as a low-level lingua franca that can be targeted by compilers. Stroustrup's original C++ implementation produced C as its output, as did Eiffel. So JS is the "C of the web".
Wednesday, July 20, 2011 1:59:18 PM UTC
Brilliant!!! Well, I'm just excited because reading through this post gave me some much needed perspective on languages such as Objective-J and CoffeeScript, as well as libraries such as GWT and WebSharper and their role in web development going forward. I hope that MS takes more deliberate steps to enter this space. I don't think jQuery (& friends) support is sufficient to compete against the likes of GWT or Google Closure Tools or Dojo or ExtJS or Sencha or SproutCore or Cappucino or... you get my drift. I had heard that Project Volta was related to this space although I'm not sure whether it went as far as providing a presentation model for developing standard interactions with the user (same thing for Script#). Too bad it seems to have disappeared off the face of the internet.
Wednesday, July 20, 2011 2:59:02 PM UTC
*Some folks think that saying "JavaScript is Assembly Language for the Web" is a totally insane statement*

I'd say that's an easily defensible statement, but saying that anyone who doesn't like ViewState is ignorant was pretty far beyond the pale.
David Fauber
Wednesday, July 20, 2011 4:22:27 PM UTC
I really enjoyed these posts. I think the evolution of the browsers have contributed greatly to the power of JavaScript. For example, IE9 is incredibly fast at executing JavaScript. If such investments were not made, would JavaScript be getting all this attention?
Wednesday, July 20, 2011 5:29:23 PM UTC
Do you trying to hint us something about Windows 8 and HTML5/JS strategy?

I like the idea of JavaScript as a assembly language. if XAML plus C# can be compiled to HTML5/JS, that will be perfect.
Wednesday, July 20, 2011 6:24:36 PM UTC
David - Touché! Gotcha.
Wednesday, July 20, 2011 6:42:18 PM UTC
I agree with Daniel's post above. JavaScript is more analogous to C than to assembly:

- Nowadays you don't code in assembly unless you need to perform expert optimisation, or use special instructions for tasks like video decompression (or you're just a masochist). C and JavaScript work equally well as both a target language and one for actual development.

- JavaScript wasn't intended as the zenith of programming language evolution, or the one developer platform to rule them all. Like C it was a product of its era, and the requirements and constraints that produced it still show in its design. Many people didn't, and still don't, like it, and there was an assumption that it would soon be replaced with a better, more perfect solution. Many people tried to create this replacement, and some of the things they developed became popular and useful, but they never achieved quite the same ubiquity, and it turned out that these supposedly perfect replacements had own problems of their own. Meanwhile, C and JavaScript stuck around, even as many of their supposed successors fell away, and people discovered a sort of beauty in their imperfection.

- Both C and JavaScript are highly cross-platform yet also platform-specific. You can find a C compiler for almost any platform, although you'll often be using a runtime that exposes platform-specific capabilities and extensions. In the same way, JavaScript is found on almost every computing device manufacturer today, sitting within a browser environment that specific capabilities and extensions.
Wednesday, July 20, 2011 7:04:22 PM UTC
I would think guys that actually wrote or write assembler don't appreciate this analogy as javascript is a very easy language to learn and you don't really have to know what you are doing to use it. In fact, you really don't need to know much about the browser or engine that runs it period.

The only thing to me that makes this analogy work is that you can't get any closer to the metal of the browser than with javascript.
Wednesday, July 20, 2011 10:59:34 PM UTC
Right _now_ Clojourescript is beeing announced - Clojoure compiled down to JS.

http://www.ustream.tv/channel/clojurenyc

Wednesday, July 20, 2011 11:37:54 PM UTC
To those who say that Javascript's lack of type safety is a downside, check out Google's Closure Tools. The Closure Compiler adds a great type system (based on the original ES4 proposal) to Javascript. The type annotations are done in JSDoc comments and are optional, but when used religiously, the compiler can do all sorts of static checking and optimizations. I have to say, writing JS directed at the Closure Compiler has been one of my favorite programming experiences.
Thursday, July 21, 2011 12:32:30 AM UTC
if js is the assembly language of the web as some ppl thought, what's the modern languages of the web which can replace the js to make programming job more easier??
Thursday, July 21, 2011 1:34:09 AM UTC
@Jay Young
I say javascript's lack of native, static type safety is a downside for performance. Clearly there are code quality wins to be made with Closure's compiler that enables one to optionally declare type annotations and also allows Closure to recompile your javascript into better-performing javascript, but this is certainly no substitution for a true static-typing system built in to the language.

Significant performance benefits can be made simply because of the static assertion that a variable can only be of a certain type and will not change over its lifetime. Javascript engines have to be coded very carefully to correctly deal with dynamic type changes in an efficient manner which is the fundamental barrier to improving javascript execution performance beyond the current state of the art. I'm sure other execution engines will come along trying out new ideas, improving performance bit-by-bit here or there, but they simply will not be able to cross this barrier.
Thursday, July 21, 2011 3:04:48 AM UTC
Not too long ago there was the "One Framework to Rule Them All" War between Prototype, jQuery, MooTools and EXT JS. Are we at a crest in the evolution of how we view and use JavaScript? No more the Framework Wars. Maybe we are in a time of Designer Languages that compile to the JavaScript VM.
Thursday, July 21, 2011 8:45:57 AM UTC
Developers should create abstractions - and many of them. A complex application doesn't fit into the mind of any "sane" person. So the problem needs to be broken down into a set of smaller problems. Just like some mathematical equations are elegantly simple, they can conceal a vast body of knowledge.

ViewState in ASP.NET is an example of such an abstraction - jQuery is another, but just because jQuery excels in DOM manipulation doesn't mean you should build your entire application in it, but certainly not without it.

More often than not I see ASP.NET developers lack a proper understanding of how ViewState works and to their chagrin find their applications bloated and slow in production albeit fast on their local machines.

Does that mean ViewState is a bad invention? Or is it merely a feat that must be undertaken to produce a well-crafted application ?
Thursday, July 21, 2011 9:17:02 AM UTC
Yes, Javascript is great, both as a language for humans and as a compilation target. And like many people, I rediscovered it through jQuery. Please keep it this way: minimal and efficient.

@Brendan: regarding CPS, the problem can be solved with callback patterns rather than state machines. I wrote a detailed post about it. It works well with preprocessors and could also be integrated into compilers but I don't know how efficient the result would be, compared to state machines.

Bruno
Thursday, July 21, 2011 10:22:39 AM UTC
@James Dunne
Even though static type information is not explicitly provided in the code, modern JavaScript engines can infer type information and use it to optimise code, falling back to slower paths only when required.
Friday, July 22, 2011 11:30:38 PM UTC
@ctrlShiftBryan, you have absolutely no idea what you're talking about.
Tom Collins
Monday, July 25, 2011 9:12:49 PM UTC
I'm curious as to why you didn't include more mention of GWT in your article. As near as I can tell, it is, by far, the strongest example of an SDK that actually treats javascript like assembly. Each browser shares the same core instruction sets of JS with each other, but if you've actually had to rewrite an entire web app to be compatible with IE's jscript, you'll know that JS is far from "Write once, run everywhere". It's not even close. Rather than deal with Win v. Mac v. Linux, it's now FF v. Chrome v. Safari v. Opera v. IE6, IE7, IE8 and IE9 {every version of IE introduces new incompatibilities with previous versions}.

I would also like to mention the acronym JSNI. Javascript Native Interface. This is a GWT term for when you write a native java method that uses javascript instead of C or Assembly to perform low-level, browser-dependent functions. You are very right that the JVM is not a good engine to become the de facto web standard, and that is why it failed... But the java language itself is a perfect example of how to bind strongly-typed OOP concepts to a weaker typed base language like JS.

I strongly disagree the quote from Nick Thompson that denounces java because: "They wrote their string handling code in an interpreted language rather than taint themselves with C!". The very fact they kept core libraries wholly virtual is why GWT can use java to smooth out browser inconsistencies in JS to do amazing things, like Porting Quake II to run in a browser http://code.google.com/p/quake2-gwt-port/ . Making a virtual language dependent on C at it's core would just make it another C lib, and not a virtual language at all.
Alyxandor James
Saturday, August 27, 2011 6:22:02 PM UTC
Great post about JavaScript. Thanks for sharing. Looking forward for more.
Sunday, August 28, 2011 10:03:42 AM UTC
Would you be more impressed if you did a View Source and found that it was not only pretty on the outside but also inside?

Actually it depends on your definition of "nice". If you mean it semantically or HTML that actually is valid, then I agree.
If you mean "nice" in terms of readable, well-formatted code, but sacrificing user experience because of document size, no.

Not so long time ago, I would have agreed with both, but nowadays Developers tools like those in Chrome take care of beautifying "ugly" HTML and JS automagically.
Tuesday, October 11, 2011 5:29:03 PM UTC
For historical context, Erik Meijer was talking about JavaScript as the assembly language of the web at a .NET language developer's mini conference at Microsoft in summer 2007. By complete coincidence, I gave a presentation at that conference on the similarities between JavaScript and x86 assembler and encouraged the language devs attending to consider JavaScript as a target platform, not just a web app scripting language (based on my immediate experience developing at scale with JavaScript, compared to previous work in Delphi, C#, and x86 assembler). Erik was delighted. ;>

Erik's Volta project (compiling MSIL to JavaScript) went public in December 2007. http://lambda-the-ultimate.org/node/2563
Comments are closed.

Disclaimer: The opinions expressed herein are my own personal opinions and do not represent my employer's view in any way.