Add KHR_draco_mesh_compression support.

[donmccurdy]

See mrdoob/three.js#13194

[zellski]

Oh I’ve been waiting for this. Do you intend to update your “public web tool” with the relevant code?

[donmccurdy]

Yep, this issue is just meant to track adding it to the web tool. The three.js PR is not going to land until after the r90 release, but we could probably merge the changes here before that.

[donmccurdy]

Staged an experimental version:

• Demo: https://gltf-viewer-experimental.donmccurdy.com/
• Code: https://github.com/donmccurdy/three-gltf-viewer/tree/feat-draco

Currently it only works for the first Draco model imported per pageview, and throws an error for the next. Not sure yet whether that's a bug in my implementation or the decoder module. Fixed.

[zellski]

It's so great to be able to FBX2glTF --draco --binary again, and just pull the (much-shrunken) result right into this viewer.

[cx20]

I tried glTF-Draco format file using experimental demo site.
https://gltf-viewer-experimental.donmccurdy.com/

• Duck.gltf Draco format model of the glTF-Sample-Models repository could be displayed.
• Duck.gltf Draco format model outputted with the latest COLLADA2GLTF could not be displayed.

When loading failed, the following error was output to the console.

bundle.js?v=r90:223 
  THREE.DRACOLoader: Decoding failed: Failed to decode geometry data.
  convertDracoGeometryTo3JS @ bundle.js?v=r90:223

The arguments used for the conversion are as follows.

COLLADA2GLTF-bin.exe -i Duck.dae -o Duck.gltf -v 2.0 -d -s

I could not judge whether it is a converter issue or a viewer problem.

[donmccurdy]

Thanks @cx20 — the COLLADA2GLTF file looks ok in Babylon and gltf.insimo.com (slightly different implementation in three.js) so i believe this a bug in my viewer or THREE.GLTFLoader. Will look into that soon.

[donmccurdy]

Released draco support. The COLLADA2GLTF issue isn't limited to my viewer, it happens in plain three.js as well, so I've opened mrdoob/three.js#13870.

[EtagiBI]

Hmm... Is it a validator problem then?

[lexaknyazev]

Yes. It means that the validator doesn't support that extension. The message will be refined in the next release.