[x3d-public] NIST conformance example behaviour

Thu Jan 26 02:46:25 PST 2017

Hi all,

Patrick, many thanks for your further comment. As Don said, we did discuss this at the X3D WG meeting yesterday. It is a difficult point to get technically correct in the standard. You are right to point out that the text in the standard needs to be precise.

It is worth reproducing the example here. It helps to consider what is going on. So, here is the listing for the XML encoding, minus the header information:

  <Scene>

    <Background groundAngle='1.57' groundColor='0 0.5 0 0 0.5 0' skyColor='0 0 1'/>

    <Viewpoint description='Front View'/>

    <NavigationInfo DEF='NAV1'/>

    <Transform>

      <TouchSensor DEF='TOUCH' description='touch to toggle light switch'/>

      <Shape>

        <Appearance>

          <Material/>

        </Appearance>

        <Box/>

      </Shape>

    </Transform>

    <Script DEF='LIGHTSWITCH' directOutput='true' url='"headlight_range.js" "http://www.web3d.org/x3d/content/examples/ConformanceNist/BindableNodes/NavigationInfo/headlight_range.js"'>

      <field accessType='initializeOnly' name='nav1' type='SFNode'>

        <NavigationInfo USE='NAV1'/>

      </field>

      <field accessType='inputOnly' name='toggle' type='SFBool'/>

    </Script>

    <ROUTE fromField='isActive' fromNode='TOUCH' toField='toggle' toNode='LIGHTSWITCH'/>

  </Scene>

Let’s consider some terminology,  as it applies to this example.

The bindable node in question is the NavigationInfo node with the DEF name “NAV1”. This node also appears as a child of the user-defined SFNode field ‘nav1’ in the Script node, where it is USEd. Of  course, as the standard 19775-1 subclause 4.4.3 (http://www.web3d.org/documents/specifications/19775-1/V3.3/Part01/concepts.html#DEFL_USESemantics) states, the USE reference is not a copy, but a reference to the original definition. So, nodes are never multiply instanced, which is contrary to what the text says in subclause 7.2.2 (http://www.web3d.org/documents/specifications/19775-1/V3.3/Part01/components/core.html#BindableChildrenNodes). This was, in fact, the reason that the Mantis issue 749 was raised, where the original proposal was to change “multiply instanced” to “multiply referenced instance”.

However, this node does appear multiple times in the scene graph. User defined fields of a Script node are treated the same as all the other built-in fields, in terms of instantiation. There is nothing special about them. So, the NavigationInfo node does appear in the scene graph as a multiply referenced node instance. When an implementation loads this file, it will create one instance of the NavigationInfo node. That instance will be referenced by whatever mechanism in the implementation is used to keep track of root nodes, and also by the Script node field ‘nav1’. We conclude, therefore, that the NavigationInfo node will appear twice in the scene graph hierarchy. This, therefore, cannot be used as the differentiator.

What about ‘traversal’. What do we mean by that? When does it occur? A scene may be traversed for multiple reasons. For example, it could be for rendering, or determination of a bounding box, or for searching, or for counting polygons, to name but a few. All these involve scene graph traversal. What happens in a Script node in these cases? This question can be answered by asking when the script is activated. Referring to subclause 29.2.2 (http://www.web3d.org/documents/specifications/19775-1/V3.3/Part01/components/scripting.html#Scriptexecution) we see that the primary cause of activation of a Script node is the receipt of an event, although it may also be activated on initialization and shutdown (Note: we won’t get bogged down by asynchronous scripts here). There is no other source of activation. So we conclude it does nothing in response to direct scene graph traversal for purposes such as rendering.

In summary, a Script node can be thought of as an event handling node. Hence, in order to enable event handling using it’s directOutput mechanisms, it must be able to reference other nodes. There is only one mechanism to accomplish this. And that is USE. So, Script nodes have to be an exception to the rule relating to multiply referencing bindable nodes. With this in mind we tried to add text to the standard to permit this exception. Perhaps, with further consideration, we can improve the presentation of this, remarking on the exception directly in the paragraph with the restriction, so that the exception immediately follows, and is associated with, the restriction.

All the best,

Roy

-----Original Message-----
From: x3d-public [mailto:x3d-public-bounces at web3d.org] On Behalf Of Don Brutzman
Sent: 26 January 2017 06:28
To: Patrick Dähne <patrick.daehne at igd.fraunhofer.de>
Cc: Andreas Plesch <andreasplesch at gmail.com>; X3D Graphics public mailing list <x3d-public at web3d.org>
Subject: Re: [x3d-public] NIST conformance example behaviour

Sincere thanks for your follow-up note Patrick.

On 1/25/2017 4:50 PM, Patrick Dähne wrote:

> Hello Don,

> 

>> "A Script node with a field containing a USE reference to a bindable node has the ability to bind a node directly by changing the value of the bind field.  Setting the bind value is equivalent to receiving a set_bind event.“

This statement was the result of 20-30 minutes careful discussion on the weekly teleconference today.

I should have also noted that it was refined and documented as part of Mantis issue 749, which was previously submitted.  Web3D Consortium Member access on the website is necessary to view these issues.

                Mantis 0000749: 7.2.2 Bindable children nodes - Amend prose

                 <http://www.web3d.org/member-only/mantis/view.php?id=749> http://www.web3d.org/member-only/mantis/view.php?id=749

It seems to me that most everything said by everyone on this topic (both email and teleconference) has essentially agreed on functionality: that a Script field USE node does not get processed as part of the scene-graph traversal, nor does it get bound simply by being referenced as an SFNode field.  We carefully discussed the specification prose today to ensure that all cases were considered.

Interestingly the participants (Dick Roy Joe Leonard Vince and myself, maybe Nicholas was still with us) looked quite closely at the exact same sentence identified by you.  We tried a few prose variations along the lines of what you are suggesting as well.  We decided eventually that the words there were still the best we could achieve.  Perhaps your phrasing is more helpful, but functionality is agreed upon at least.

So, for your question below, why do we want to add this additional sentence quoted at the top of this message?  Continued review of paragraph 7.2.2 steps (a) through (i) describes binding behaviors on the basis of events being received.  The case of a Script node with a field USE reference and directOutput='true' which subsequently sets the value of 'bind' field during Script invocation wasn't addressed... so this extra sentence was added for clarity, to explicitly note that the bind/unbind operations are equivalent for event access and direct access.

I hope this helps explain the additional sentence.

Of note is that the in-person discussions are really helpful. Interpreting the specification prose can be a tricky business!  We do our best to listen to every option together and achieve concise descriptions.  We also reference the email thread and every direct suggestion in Mantis for future reference as well.  Also essential is the presence of a good example scene with agreed-upon semantics for rendering and user interaction.

Since Fraunhofer is a member you are of course welcome to join.  If you or anyone else wants to propose a further improvement, we will be happy to schedule a good time for that.  Member connection information is online, non-member invitations are not a problem for these topics.

                 <http://www.web3d.org/member/teleconference-information> http://www.web3d.org/member/teleconference-information

So... lots of careful listening and due diligence.  We are further paying close attention to using Mantis when performing the corresponding X3D specification editing in Github.

Please check out the mantis issue if you still have concerns, and let's review again at a good time for you (and Andreas as well).  There is little value in "winning" an email debate, rather we always want to achieve common ground, shared understanding and (hopefully!) consensus so that all implementations can consistently handle X3D content.  So these efforts are all worthwhile, especially for the special semantics of bindable nodes, Script execution and DEF/USE referencing.

More specification issues await... but we are certainly operating at a greater level of working group + community knowledge than ever before.  Apparently benefits are cumulative!  8)

Hope we're looking ~ acceptable now.  Further improvement always welcome.  Looking forward to continuing progress together with all of X3D, again thanks to all involved.

v/r Don

> Sorry, I did not want to resume that nitpicking discussion, but you are taking the wrong direction :-(

> 

> My point is:

> 

> 1. There is this sentence in section 7.2.2 of the specification: „The results are undefined if a multiply instanced (DEF/USE) bindable node is bound.“

> 

> 2. In the NIST example we exactly do have that case - a bindable node (DEF nav1 NavigationInfo) that is multiply instanced - once as a child of the (implicit) root Group node, and once as a child of the Script node.

> 

> 3. => According to the specification, the example is undefined. Q.E.D.

> 

> I do not claim that there *actually* is a problem with the NIST example, I just say that *formally* the example is invalid because of that sentence in section 7.2.2 of the specification.

> 

> So what is the rationale behind that sentence? That is explained in the Annotated VRML97 Reference. The browser needs to determine the transformation of the currently bound Viewpoint, Background or Fog node. The NavigationInfo node actually is a special case, because for that node the transformation of the currently bound Viewpoint is taken. Now, when you DEF/USE a bindable node (or any group node above a bindable node), you have to take care that you do not add the bindable node more than once to the transformation hierarchy, because otherwise the browser cannot decide which of the transformations to take. Adding a bindable node to the Script node does *not* add it to the transformation hierarchy (see section 4.3.5 of the specification) - that is why there is no problem in the NIST example.

> 

> So, what to do with that awkward sentence in section 7.2.2? My recommendation would be to replace it with a sentences like this: „A bindable node may only be bound when it is part of the transformation hierarchy exactly once, otherwise the result is undefined.“ But I’m sure a native english speaker can word that much better.

> 

> Bye,

> 

> Patrick

all the best, Don

-- 

Don Brutzman  Naval Postgraduate School, Code USW/Br        <mailto:brutzman at nps.edu> brutzman at nps.edu

Watkins 270,  MOVES Institute, Monterey CA 93943-5000 USA   +1.831.656.2149

X3D graphics, virtual worlds, navy robotics  <http://faculty.nps.edu/brutzman> http://faculty.nps.edu/brutzman

_______________________________________________

x3d-public mailing list

 <mailto:x3d-public at web3d.org> x3d-public at web3d.org

 <http://web3d.org/mailman/listinfo/x3d-public_web3d.org> http://web3d.org/mailman/listinfo/x3d-public_web3d.org

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://web3d.org/pipermail/x3d-public_web3d.org/attachments/20170126/c7a12bb0/attachment-0001.html>