Re: Comment on Device Orientation API

Hello,

We discussed this issue at the face-to-face meeting today. Our
conclusion was that it is very difficult to provide normative language
that would force implementors to guarantee that device orientation
events are generated for changes that are over a specific angular
threshold.

However, we do understand that the importance of ensuring that
implementations of this spec are as similar as possible. For this
reason, we will add non-normative language that recommends a maximum
angular threshold of 1 degree.

Thanks,
Andrei

On Thu, Mar 1, 2012 at 11:03 AM, Wolfgang Damm
<wolfgang.damm@wikitude.com> wrote:
> Hi all,
>
> my name is Wolfgang Damm and I'm working at Wikitude as Lead Software
> Architect. In that role I'm responsible for developing an AR prototype using
> the DeviceOrientation API. Let me jump right into the discussion.
>
> Lars, thanks for adjusting the cap.
>
> But, I still think that the specification needs to be reworded. The
> pharse "significant change in orientation" is to vague and can be
> interpreted in various ways. Even if it is problematic for our
> usecase, Opera was still well within the specification.
>
> A few usecase come to my mind (and I think you have thought of a few more
> already):
> 1. Augmented Reality
> 2. Games
> 3. Determining Screen orientation (only Up, Down, left, right is of
> interest)
>
> With the current specification I believe only usecase 3. is sufficiently
> covered. For 1. and 2. it is up to the specific UA implementation if they
> are possible.
>
> In my opinion, to cover the different update rate/accuracy needs of
> different usecases without draining battery or degrading performance the
> optimal way would be to let the web app specify how many updates/what
> accuracy it needs. But if this is not an option, there should be a more
> concrete wording of when events should be fired. Something in the line of:
>
> The event should be fired with a minimum update rate of around 100 ms or if
> a significant change in orientation occurs.. The definition of a significant
> change in this context is left to the implementation.
>
> By defining the update rate, a game developer knows what to expect when
> using this API. It also helps in development that the event firing behavior
> is more similar across different UAs..
>
> Best,
>
> Wolfgang
>
> 2012/3/1 Lars Erik Bolstad <lbolstad@opera.com>
>>
>> Hi Martin,
>>
>> Sounds like we've capped this at a too high level in Opera's
>> implementation. We'll adjust that.
>>
>> Lars Erik
>>
>>
>> On 29.02.2012 21:40, Doug Turner wrote:
>>>
>>> Hi Martin,
>>>
>>> Thanks for your suggestion.  This is an interesting idea, and one
>>> that was discussed.  The draft states:
>>>
>>> The event should fire whenever a significant change in orientation
>>> occurs. The definition of a significant change in this context is
>>> left to the implementation. Implementations may also fire the event
>>> if they have reason to believe that the page does not have
>>> sufficiently fresh data.
>>>
>>> I tend to think that Opera filtering at 4 degrees is a UA bug.  Maybe
>>> a RFE.  I do not think I want to expose tighter control here to web
>>> application.  Surely any change to the API allowing such control
>>> could be veto'ed by the UA.  Opera could still cap their
>>> notifications at 4 degree changes even though the web application
>>> asked for something tighter.
>>>
>>> It could also be an implementation concern.  I know when I fire lots
>>> of events, the over all browser performance suffers.  I think UA's
>>> need some flexibility here to provide the best possible performance.
>>>
>>> Thanks again,
>>>
>>> Doug Turner
>>>
>>>
>>>
>>> On Feb 29, 2012, at 6:46 AM, Martin Lechner wrote:
>>>
>>>> Dear members of the Geolocation Working Group,
>>>>
>>>> I am Martin Lechner, CTO of Wikitude, a member of W3C and developer
>>>> of the Wikitude World Browser.
>>>>
>>>> We are currently developing an augmented reality project on
>>>> platforms based on the DeviceOrientation API, and do have a comment
>>>> regarding some specific items in the draft specification.
>>>>
>>>> The current specified update rate of the device orientation event
>>>> is not sufficient for our usecase. Specifically the phrase: “the
>>>> event should fire whenever a significant change in orientation
>>>> occurs.” leaves it up to the implementation when exactly to fire an
>>>> event.
>>>>
>>>> We are already seeing different implementation in the prerelease
>>>> versions of Opera (LAB Opera Mobile) and Mozilla (Aurora). While
>>>> Mozilla reports events quite frequently (which is preferred by us),
>>>> Opera matches the specification and reports events only if there is
>>>> a change of approximately 4 degrees.
>>>>
>>>> Our suggestion would be to let the web developer define how
>>>> accurate the orientation changes should be reported. E.g. like the
>>>> accuracy modes in Geolocation API. For AR it would be nice to have
>>>> a high accuracy to achieve more events to get a smoother, more
>>>> accurate calculation when moving the device.
>>>>
>>>> Best regards, Martin Lechner
>>>>
>>>> -- - - - Martin Lechner CTO
>>>>
>>>> Wikitude GmbH Ginzkeyplatz 11 5020 Salzburg/Austria Phone +43 662
>>>> 243310 Mobile +43 676 840 856 300
>>>>
>>>> http://www.wikitude.com
>>>>
>>>>
>>>>
>>>
>>
>>
>
>
>
> --
> Wolfgang Damm
> Lead Software Architect
>
>
> Wikitude GmbH
> Ginzkeyplatz 11
> 5020 Salzburg/Austria
> Email: wolfgang.damm@wikitude.com
> Phone: +43 662 243310
> Mobile: +43 676 840 856 900
> Web: http://www.wikitude.com
> Twitter: twitter.com/Wikitude
> Facebook: facebook.com/wikitude
>
> Make sure to subscribe to our developer newsletter so you don't miss any
> update about ARchitect.
>

Received on Monday, 5 March 2012 14:27:12 UTC