Fix sample models

[lasalvavida]

This should resolve #556, #571, #524, and #517.

[pjcozzi]

Replaces #564.

[pjcozzi]

@TomPed would you like to test this by dragging and dropping each model (including each format variation in the sub-directories) onto http://cesiumjs.org/convertmodel.html. I think the ones with separate resource files (.bin, images, shaders) will need to be tested explicitly in Sandcastle.

Also validate each one with http://lexaknyazev.github.io/gltf-demo/

[pjcozzi]

"bSample" should be "Sample"

[pjcozzi]

Also, up to you, but it would not be a bad idea to replace each README.txt with README.md for consistency:

See the README.txt in each model's directory for usage restrictions.

Maybe even include the thumbnail and short description here also in the README.md for each model.

[pjcozzi]

Also update this .md:

A README.txt with any usage restrictions.

[lasalvavida]

Good catch. And I think that's a good idea for the READMEs, I'll add that when I get a chance.

[pjcozzi]

I love the new animated gifs for the animated/skinned models!

[pjcozzi]

The issues in #556 look like they have been addressed.

Just waiting on @TomPed's testing and any additional feedback from @lexaknyazev before merging.

[TomPed]

each model (including each format variation in the sub-directories) onto http://cesiumjs.org/convertmodel.html.

@pjcozzi I only see one type of COLLADA file to send to the converter Never mind.

[TomPed]

I am getting some warnings about Unexpected propertys when I send some .gltf files into the glTF Validator. Is this a problem?

[lasalvavida]

@TomPed it depends which versions you're talking about. glTF and glTF-Embedded should be totally clean. glTF-MaterialsCommon has some warnings like:

[WARNING] glTF: materials/Effect-inner/extensions/KHR_materials_common: Unexpected property doubleSided

from the validator. These are issues with the validator (and the spec for the extension which needs some updating and clarification).

glTF-Binary has the following:

[SEVERE] glTF: buffers/binary_glTF/uri: Unsupported MIME type: text/plain

As best I can tell, this is also a validator issue.

[TomPed]

Same. I see the same results. I've just ran all the models.

[lexaknyazev]

KHR_materials_common files have 2 kinds of issues:

1. Unexpected properties (doubleSided, jointCount and transparent) in the extension object. Schema says (maybe wrongly) they must be inside values.
2. Textures for ambient property in VC.gltf. Schema permits only FLOAT_VEC4.

Those MIME errors come from "uri": "data:," fields because data: URI without explicit MIME type implies text/plain which is incorrect type for buffers or images.
I'll remove this validation on empty data URIs.

/ReciprocatingSaw/glTF-Embedded/ReciprocatingSaw.gltf is an empty file.

[lasalvavida]

The KHR_materials_common spec hasn't been ratified, and does need some attention. I'm inclined to follow the implementation and update the spec to match, but that's probably up to @pjcozzi.

Thanks @lexaknyazev! Good catch on the empty file

[pjcozzi]

@lasalvavida for KHR_materials_common, I agree, update the spec.

• See Common Materials Extension #424 (comment)
  • It would be worth going through that entire issue and seeing if anything else in the extension spec needs to be updated.
• Are both COLLADA2GLTF and Cesium in-sync?

[TomPed]

I think the ones with separate resource files (.bin, images, shaders) will need to be tested explicitly in Sandcastle.

You are right. I've tested all other models and they work. I will now be testing these types of models.

[lexaknyazev]

Validator's MIME type checks fixed.

[TomPed]

Alright so I've tested everything. All looks good but a few issues.

1. sampleModels/BoxSemantics/glTF-MaterialsCommon/CesiumLogoFlat.jpg is missing
   • The file names in sampleModels/BoxSemantics/glTF-MaterialsCommon/ should also be changed to start with a capital letter for consistency
2. Origin of GearboxAssy is off (for all types of models)
3. Origin of Monster is off (for all types of models)

[lasalvavida]

@TomPed

1. Good catch! Names are fixed and the missing image is added
   2/3. I think that's just how the models are. Choosing the origin for a model is a somewhat subjective topic, and as far as I'm aware, there's not a really a correct choice.

[pjcozzi]

@lasalvavida @TomPed is anything else needed before merging this?

[TomPed]

From me no, all formats worked.

[pjcozzi]

Thanks so much @lasalvavida @TomPed @lexaknyazev! This is a great contribution.