Web Browser Engineering (2021)

188 points by ibobev 22 hours ago on hackernews | 27 comments

sublinear | 22 hours ago

> On top of all of that, modern browsers also execute WebAssembly, a hardware-independent bytecode format for many other programming languages to target, and which may one day be co-equal to JavaScript on the web.

I skipped around a bit, but when I stumbled onto this opinion I backed out.

This is not the point of wasm at all. Wasm is there as a compatibility layer and for squeezing extra performance out of critical sections of code. The decisions to restrict its API access are very sensible. We can't honestly still be confused about this in 2026.

I excused the python code since a serious browser is not the point, but I'm now suspicious of other intuition-building done throughout the rest of the book.

suriyaG | 21 hours ago

what do you find about this opinion to be a problem?

I've worked with wasm and am fairly familiar with the ecosystem. it sounds like it captures the essence quite well

pavpanchekha | 21 hours ago

Hi, author here! I made a real effort to keep opinions out of the book, and I'm wondering what you feel is opinion above! I don't mean to imply that WebAssembly either will or should be co-equal to JavaScript, just that 1) it's not covered in the book at all, and 2) who knows what the future will contain.

In any case, this line is in the appendix describing what we didn't cover and, per that, we don't say anything else about WebAssembly in the book!

enos_feedler | 21 hours ago

Perhaps nix the part about it being co-equal to JavaScript in the future. Im not sure its even a goal of wasm to get there and it just invites grumbles like parent over there

throawayonthe | 20 hours ago

sublinear | 20 hours ago

Not a link to a W3C decision. There have been drafts. They stayed drafts for a reason.
Hey, just imagine: Rustact, a React-a-like where you write the code in Rust and everything gets compiled to a massive blob of WASM to be rendered client-side. It's completely inaccessible! I mean, it's completely non-discoverable! I mean, it's completely UI/UX Compliant because it will be Mystery Meat Navigation from The UI/UX Experts, and it will be so much more performant to shove a 100 MB blob down the line to everyone than have 10 10 kB blobs being served on-demand.

You know the people who thought Java Applets, Flash, and Sliverlight were Wonderful Ideas would be on top of that one.

CharlesW | 20 hours ago

> It's completely inaccessible! I mean, it's completely non-discoverable!

Solutions like this can still render to DOM instead of Canvas. Here's Rustact, but with Swift/Swift UI: https://elementary.codes/

Why do you want to limit Wasm to what it's "supposed" to do?

Direct DOM access would make it possible to build full sites in Rust or C. There is no real security tradeoff and very little complexity (expose an existing API to an existing runtime). In return you get a web that's faster and completely open in terms of languages.

Right now the only options are use canvas and draw everything from scratch, or have a JS layer that transfers messages between the code and browser. You have to choose bloated size and zero accessibility, or pure overhead from copying data back and forth.

pavpanchekha | 21 hours ago

Hi folks, one of the authors here. Happy to answer questions. Chris and I wrote the book starting in 2019, and the book was largely done in late 2024, with publication in 2025. In case folks are wondering, literally none of it was AI-written, because the AIs didn't exist yet!

callc | 20 hours ago

Hey Pavel! Thank you so much to you and Chris for writing this book. I really mean it! It's started me on a fun journey (and hopefully career change soon) into working on really hard problems in browsers.

I finished going through this in July this year. Here are my thoughts at the time: https://calvinlc.com/p/2026/07/23/browser.html

No questions, just gratitude!

pavpanchekha | 20 hours ago

Great write up, thanks for sharing!

shoobiedoo | 19 hours ago

Amazing work. I've always heard that browsers are more complex than operating systems but I never thought to investigate how that could be. Your book shows the layers beautifully

jamesfinlayson | 18 hours ago

Good to see it finished! I must have seen this on HackerNews last year and scribbled it on a post-it note of various tech books I need to buy and read some day.

12purple | 9 hours ago

I would be very interested in this list of books :-)

i_am_a_peasant | 5 hours ago

ah excellent, just bought it on the kindle, thank you for this great contribution to public knowledge!

TZubiri | 21 hours ago

pangram says not LLM

Congratulations on not slopping out

pavpanchekha | 21 hours ago

The book was written between 2019 and 2024, and published in 2025!

JimDabell | 9 hours ago

Not every conversation needs to be a conversation about AI.

coreyp_1 | 20 hours ago

Nice.

I'm on the path to building a browser. I had vaguely remembered this site from when it was listed before, but I couldn't find it when looking again. It's bookmarked now. :)

I'm writing my own libraries (because, with AI, why not indulge in low-level fun?), and I'm almost to the layout engine.

So far (all C libraries), I can read multiple image formats, parse HTML/XML and CSS, rasterize text and just about anything that can be required by CSS or SVG. Fonts were fun... I can do TTF but not WOFF yet. I can do some WASM, but no JS engine yet. I do have my Regex engine, though. (I had built WASM first because it's being used for another project). Still a long way to go. Only approx 780K lines spread out across 30 or so libraries written so far.

pavpanchekha | 20 hours ago

That sounds incredibly fun. Hope you enjoy the book!

m00dy | 11 hours ago

I’d be happy to share what I know about browser fingerprinting.

Email: eren@evenstamp.com

socalgal2 | 10 hours ago

> all C libraries

I'm sure you have your reasons but this sounds like a disaster waiting to happen. Jonathan Blow might be able to never write memory bugs but something as big as a browser will likely, eventually have more than you edit it, or the you of the future is not the same you of today and so will forget all the issues that C is not preventing you making.

A browser needs to be one of the most hardened pieces of software there is since it runs untrusted code. All the browser makers are trying to figure out ways out of their C++ legacy use-after-free exploit hell. Servo in rust, LadyBird going rust, WebKit considering switching to all Swift, Chrome converting portions to rust.

kuharich | 19 hours ago

NooneAtAll5 | 18 hours ago

who tf thought .engineering should be allowed to be a TLD?

pavpanchekha | 18 hours ago

Author here. I assume .engineering was intended for, like, civil engineering firms. But, fun fact, I named the book by going through the list of TLDs since, after all, a book about web browser needs a URL about as much as it needs a title. The URL web.browser.engineering is supported.

lopis | 10 hours ago

Maybe it sounds silly, but is there any way to read this book in paper? I can't focus on very long texts on the computer these days.

edit: never mind, found it for sale :)