Mozilla Servo: the road to Acid2 | Hacker News Reader

Mozilla Servo: the road to Acid2

groups.google.com

154 pointsbpierre16 comments

"Now that we've essentially passed Acid1 (once one remaining pull request is merged), Acid2 is on the horizon!"

I applaud the Servo team for reaching such a concrete milestone on the long path to implementing a modern layout engine (in an experimental and highly unstable programming language, no less). I do want to point out an amusing anecdote, though. As the Acid1 test states:

"All 100%-conformant CSS1 agents should be able to render the document elements above this paragraph indistinguishably (to the pixel) from this reference rendering, (except font rasterization and form widgets)."

The point of the parenthetical disclaimer at the end was probably to account for differences in appearance in native platform elements, but the language is weaselly enough that Servo still manages to qualify as compliant despite the fact that it doesn't have form elements at all yet. :)

http://www.reddit.com/r/rust/comments/1kz0ue/servo_passes_ac...

To be fair, I don't think the point of Acid1 was to show off form elements. Floats, percentage widths, and how the flows are constructed were the difficult bits.

Note that we tried to support everything else even if the language allowed for us to skip it. For example, line-height is used in a few places, but we waited until we had line-height support before we considered ourselves passing.

For anyone else that didn't know: Servo is a prototype web browser engine written in the Rust language.

tiny correction, it is an engine.

Currently you cant use gecko like you can use webkit, servo is meant to change that.

What's it for? Just for fun? Or do they have a particular use case in mind?

Highly parallel web rendering. Also, easier maintainability versus a comparable C++ codebase.

See https://github.com/mozilla/servo/wiki/Design for some notes on rationale.

So, is this because they're microtargetting Acid compliance? Or do they actually have 90% of a browser layout engine?

And if it's the latter, what would actually need to be done to build the rest of a browser around it?

IIRC Acid1 only targets basic CSS compliance (http://en.wikipedia.org/wiki/Acid1).

I guess that most of the DOM API (end user level) has yet to be done along with the javascript API (binding) controlling it (e.g. XHR, Websocket, DOM manipulation, domain security control, etc...).

Next : the browser itself (chrome, modules, tabs, external interfaces, UI, ...).

So, no, not 90%. But I would attribute 50% for starting the project itself and having created something as awesome as Rust.

> I guess that most of the DOM API (end user level) needs to be done along with the javascript API controlling it (e.g. XHR, Websocket, DOM manipulation, domain security control, etc...).

We do have JavaScript and DOM bindings but they're incomplete. (None of the items you mentioned are implemented, except for basic DOM manipulation.)

  > Next : the browser itself (chrome, modules, tabs, 
  > external interfaces, UI, ...).

I'd like to note that Servo itself is intended to be a layout engine independent of any specific front-end. The forthcoming UI reference implementation, a.k.a. Crow (seeing a theme yet?), can be tracked here: https://github.com/mozilla/servo/issues/111

Both are references to thing that can be associated with rust - Servo, Crow(bar), as well as references to MST3k?

But doesn't Crow sound too much like Chro[me] and lead to confusion?

As written in OP, Servo currently lacks table layout and absolute positioning. Those are pretty big items, so no, not 90%.

Acid2-passing layout engine doesn't need to be that big. http://weasyprint.org/ is known to pass Acid2 and renders the web pretty well. It does all layout by itself in less than 15,000 lines of Python.

Thanks for the pointer. I love it that the output of the test is just a smiley face.

There is a guide explaining how the smiley is drawn. Interesting reading.

http://www.webstandards.org/action/acid2/guide/

Reply on news.ycombinator.com