Live data from Hacker News

An entire Mobile OS built using HTML, CSS and JS - by Mozilla

mobile.dzone.com

111–116 of 116 posts

Re: An entire Mobile OS built using HTML, CSS and JS - by Mozilla

#111

Earlier quoted context omitted.

there is something fundamentally different, you can choose not to update a native appp, you can also choose to run an old verion. you can also continue a native app if the creator decides its not worthwhile, or google buys them out chrome is the exception not the rule

My point is that there is nothing* preventing any native app from including code to update itself. Over time, I believe that more and more native apps will do this. For example, Firefox is moving in that direction now. * Except for in the case of iOS. Apple has a policy that apps are not allowed to update themselves. But there's still no technical limitation.

Right so we're discussing the reality of the situation, which is that web apps would have no choice but to force update (as they would be run server-side and hence the client wouldn't have a choice) and that native apps could be possibly, but aren't.

Re: An entire Mobile OS built using HTML, CSS and JS - by Mozilla

#112

Earlier quoted context omitted.

>This makes absolutely no sense. This is borderline an insult to system developers -- because system software is what the device is running directly. It's ok. We're used to it. From the desktop to the laptop to the netbook to the phone, Mozilla and Google have been trying to exterminate us for years.

> Mozilla and Google have been trying to exterminate us for years. I highly recommend you take a look at PNaCl: portable LLVM binaries run in the NaCl sandbox. Since LLVM bitcode can be JITed (or even offline compiled before execution), you would be looking at optimized code for whichever target you're running on. Very exciting stuff. http://www.chromium.org/nativeclient/pnacl/building-and-test...

Which requires that people use Chrome, running a web-browser to run native code that gets JITed from bitcode... at that point, why not just pull the blocks from the Jenga tower and run native code on your processor through an executable file loaded by an operating system?

Re: An entire Mobile OS built using HTML, CSS and JS - by Mozilla

#113

Earlier quoted context omitted.

> Mozilla and Google have been trying to exterminate us for years. I highly recommend you take a look at PNaCl: portable LLVM binaries run in the NaCl sandbox. Since LLVM bitcode can be JITed (or even offline compiled before execution), you would be looking at optimized code for whichever target you're running on. Very exciting stuff. http://www.chromium.org/nativeclient/pnacl/building-and-test...

Which requires that people use Chrome, running a web-browser to run native code that gets JITed from bitcode... at that point, why not just pull the blocks from the Jenga tower and run native code on your processor through an executable file loaded by an operating system?

Because installing software sucks, this is auto-updated by default and will always be optimized for your processor.

Re: An entire Mobile OS built using HTML, CSS and JS - by Mozilla

#114

This runs a Linux kernel with some components in common with Android like libusb, so really linux is the operating system and there is a browser running on top of it. Lots more info: https://wiki.mozilla.org/B2G

Linux isn't an operating system, it's a kernel, the OS is Boot-to-Gecko, in the same way that Android is an OS with a Linux kernel.

You're right, in fact the Linux Kernel was used to minimize changes OEMs would have to make to support B2G. Also, Mozilla has been working with Qualcomm to make sure that B2G is optimized for their chips. It should be very interesting to see the potential market shift once B2G is released.

Re: An entire Mobile OS built using HTML, CSS and JS - by Mozilla

#115
post #36

I feel like I have just seen the future and it is very bright. This would make mobile development so much easier and open up a completely new range of possibilities. Just imagine .... - writing just one app that runs on every desktop and mobile phone out there. - apps which update entirely automatically, without any extra line of code, because they are basically just cached versions of a website. - a designer, who ju…

>However, I am a little bit skeptical that a web-based OS will perform the same as a native OS...

I believe the idea is that it will be native on ALL devices, thereby eliminating the need to convert one application for multiple devices. An application written in HTML5/CSS/JavaScript for the web should function the same way on a mobile device OS that is essentially a web browser. Am I wrong?

Re: An entire Mobile OS built using HTML, CSS and JS - by Mozilla

#116
post #32
post #24

Earlier quoted context omitted.

Regarding the Angry birds argument, if there were some type of deep-reaching JS "extensions" that expose lower level stuff I'd wager that you could get the game running pretty fast. I think Appcelerator/PhoneGap offer things like this. Granted, It is "twisting" JS, but I'd assume they are considering the limitations of the current "HTML5" platform and hopefully offer some solutions to get a "Web" Angry Birds version…

That's why I used the comparison, it's definitely doable and I would assume that you'd get a decent enough performance out of it (as opposed to e.g. a JS-based Rage). I just don't get the argument at all that this would enable it to run on a cheaper platform, i.e. somewhere where a native app wouldn't run. You're slicing away a tiny bit of overhead (once a kernel+browser is loaded), but adding more than that.

Wrong. Think of the three big mouths to feed in Android running modern apps or a browser (all of which require the embedded WebKit):

* Kernel

* Dalvik

* WebKit

The middle one is in fact a big memory and CPU pig for real workloads. Take it out, and B2G fits with room to spare on the 256MB QualComm phone we demo'ed at MWC -- ~128M used by kernel, Gecko, Gaia and apps at most.

Android does not even fit on this phone.

Your assumption that native-app framework and VM code loaded in practice is a "tiny bit of overhead" is false.

Post reply on HN