News for hackers. Stuff that matters.
Re:Servo Continues Pushing Forward by higherpurpose on Friday May 01, 2015 @02:38PM (#9473589)
I have a feeling Servo will finally make me give up Chrome. From what I've seen so far the performance is very impressive [1]. I just wonder if Elecrolysis will ever arrive. It seems like I've been waiting for it forever. Chrome's sandboxing is great for security and performance purposes and I can't give that up (I never seem to have a problem switching to another tab even when I have 40 of them open in Chrome, while in Firefox the whole browser "hangs" sometimes when I load a heavy tab).

[1] - https://www.phoronix.com/scan.php?page=news_item&px=MTgzNDA

Re:Servo Continues Pushing Forward by chimeracoder on Friday May 01, 2015 @02:45PM (#9473647)
> I just wonder if Elecrolysis will ever arrive. It seems like I've been waiting for it forever.

It's here already, in Nightly: https://nightly.mozilla.org/ (more info: https://wiki.mozilla.org/Electrolysis).

I haven't really had the problem you're talking about in Firefox as it is, but you can try it out and see if Nightly is any better. Keep in mind that Nightly is slower overall than Firefox, because it has a number of debugging flags enabled that impact performance.

Obligatory disclaimer: Nightly has lots of bugs and quirks compared to Firefox, so it should not be considered indicative of the normal Firefox experience[0]. Also, many extensions have not yet been ported (which is why it's available in Nightly for developers before hitting Firefox).

[0] That said, it's pretty damn stable. I'd be using it for my daily browsing except that I can't live without Vimperator, which hasn't yet been ported to e10s: https://github.com/vimperator/vimperator-labs/issues/211

Re:Servo Continues Pushing Forward by higherpurpose on Friday May 01, 2015 @03:00PM (#9473740)
I know it's in Nightly, but wasn't it in Nightly 3-5 months ago, too? Shouldn't it have moved to the "next" development version by now, like Aurora or whichever is the next "more stable" one, if Firefox follows a 6-weeks development cycle?
Re:Servo Continues Pushing Forward by callahad on Friday May 01, 2015 @03:10PM (#9473823)
Not every feature can be completed in a single six-week train, especially one as complex as rewriting huge chunks of a browser that was not initially designed with process separation in mind. :)

That said, we should have a go/no-go on Monday for uplift with Firefox 40. We're getting Very Close Now. :)

Re:Servo Continues Pushing Forward by mconley_ on Friday May 01, 2015 @04:23PM (#9474265)
If we do the uplift, I believe the plan is also to hold there until we clear out a bunch of less-critical, but shipping-blocking bugs. So it might be on Aurora for a few releases.
Re:Servo Continues Pushing Forward by simcop2387 on Friday May 01, 2015 @02:42PM (#9473620)
Electrolysis isn't a Servo thing but a Firefox thing. Something that many of us have been waiting for also. From what I've seen with the way Servo works, it should be working in basically the same way from the beginning but I'm not as confident on that as I'd like to be. Since it renders and works on each iframe in another thread and each tab is implemented as an iframe with mozbrowser extensions I would think it should fix that problem.
Re:Servo Continues Pushing Forward by Manishearth on Friday May 01, 2015 @03:08PM (#9473800)
I think he was talking about Firefox.

FWIW in Servo when talking of multiprocess stuff we colloquially call it "Electrolysis". I think Patrick has some work in progress on this. It's not as much work as the Firefox e10s (we already use threads and channels for everything), but I think there is some (relatively) minor work to be done.

Re:Servo Continues Pushing Forward by zobzu on Friday May 01, 2015 @07:08PM (#9475073)
how do you handle component sandboxing in Servo compared to firefox e10s? (firefox being child process sandbox, ipc to parent, and things being a little "unperfect" in the separation)
Re:Servo Continues Pushing Forward by pcwalton on Friday May 01, 2015 @09:03PM (#9475430)
The plan is to use gaol [1] for sandboxing. For IPC, we want to send Rust objects serialized via serde [2] over native OS IPC pipes (AF_UNIX sockets on Unix).

[1]: https://github.com/pcwalton/gaol

[2]: https://github.com/serde-rs/serde

Re:Servo Continues Pushing Forward by zobzu on Saturday May 02, 2015 @03:40AM (#9476251)
that looks nice i hope the separation between processes themselves will be nicer than firefox (or even chrome) since its "from scratch" :)

(ie that it wont be remoting everything and only protect the system from escaping a child - makes sense in firefox/chrome, less with rust being the compiler i guess)

I suspect that if servo doesnt try to be an OS like firefox/chrome do, a whole process sandbox may even be fine enough. But if it does.. :)

Re:Servo Continues Pushing Forward by gbersac on Friday May 01, 2015 @04:34PM (#9474345)
It does look like amazing performance. But It may be too early too conclude on that point. The gain of performance may only be because of less feature implemented in servo leading to less thing to compute => less computational time.
Re:Servo Continues Pushing Forward by sambe on Saturday May 02, 2015 @03:27AM (#9476222)
I never experience that in Firefox and have much more than 40 tabs open. Granted if the tab is not actually loaded, then it will go off and load it. How does Chrome's model help here?