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 working. Additionally, there is already this:
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.
That looks like a full fledged game running in a canvas. It looks great to me. Is it somehow inferior to native Angry Birds? (I've never looked any other version).
Comments
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 working. Additionally, there is already this:
http://chrome.angrybirds.com/
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.
Well sure, but how is the Web 2.0 bubble supposed to keep going without trying to displace the entire actual computer?
Someday they'll try and download the hardware as FPGA over a RESTful web API, too.
That looks like a full fledged game running in a canvas. It looks great to me. Is it somehow inferior to native Angry Birds? (I've never looked any other version).