Re: Measuring the gap between the spec and browser implementations

Hi Rich,

I'm writing a longer (I can hear everyone moan now 8)) reply to your 
message as there's a lot to consider - especially in the points you 
included at the bottom about screen orientation.

But in the meantime I thought I'd send a quick update on our dev/testing.

After some very helpful offline feedback from Tim Volodine we've updated 
our demo app to be fully spec compliant now and it works perfectly in 
Chrome on Android (M33 - currently Chrome Beta in the Play Store).

Code is on github and you can try the demo for yourself here 
http://buildar.com/static/tmp/awe.js/examples/geo_ar/index.html

Rich, I also tested this on Opera and Opera Beta and they both appear to 
work correctly - however the data is really noisy and the AR projections 
bounce around a lot. Chrome M32 (current stable version) has the same 
problem. I think Tim said they moved to ROTATION_VECTOR as the 
underlying sensor.

We've also got a code fork in there so the Firefox version still works - 
even though it's not really spec compliant (mostly to do with not 
obeying the Right Hand Rule).

But the long and short of it is that as soon as Chrome M33 goes live 
then the default Android web experience will be spec compliant and run 
Augmented Web apps really nicely 8)

However, we'd really like to contribute a bunch of diagrams and updated 
examples for the spec that I think would help a lot of developers.

And I still think the calibration discussion is really important to 
continue so we can get all the browsers to this point.

roBman

Received on Thursday, 23 January 2014 04:05:12 UTC