Add support for stored tangents and add mesh attributes table #901
Conversation
specification/2.0/README.md
Outdated
| |`TANGENT`|`"VEC4"`|`5126` (FLOAT)|XYZW vertex tangents where the w component is a signed float (-1, +1) indicating handedness of the tangent basis| | ||
| |`TEXCOORD_0`|`"VEC2"`|`5126` (FLOAT)|UV texture coordinates for the first set| | ||
| |`TEXCOORD_1`|`"VEC2"`|`5126` (FLOAT)|UV texture coordinates for the second set| | ||
| |`COLOR_0`|`"VEC3"`<br>`"VEC4"`|`5126` (FLOAT)|RGB or RGBA vertex color| |
There was a problem hiding this comment.
Normalized UNSIGNED_BYTE should be also allowed here.
There was a problem hiding this comment.
Okay, I can add that.
|
|
||
| |Name|Accessor Type(s)|Component Type(s)|Description| | ||
| |----|----------------|-----------------|-----------| | ||
| |`POSITION`|`"VEC3"`|`5126` (FLOAT)|XYZ vertex positions| |
There was a problem hiding this comment.
Why do we need to define component type here?
WEB3D_quantized_accessors requires normalized UNSIGNED_SHORT.
In general, it's not important spec-wise what the data type is as long as it matches actually stored data. Shaders shouldn't be affected anyway.
There was a problem hiding this comment.
The issue is that runtime implementations that have an abstraction for vertex data don't directly deal with buffers. BabylonJS, for example, can only accept number[] or Float32Array for uvs and vertex colors. I don't think we should leave it open ended.
Edited: After more thinking, revised wording so that it makes more sense.
There was a problem hiding this comment.
In general, runtime should be able to read any vertex format using info from accessor's fields. I agree that for the core spec other types for vertex positions make little sense.
Can we say that required extensions (e.g. WEB3D_quantized_accessors) may extend this table?
There was a problem hiding this comment.
I would say for any extension not just required extension may extend this table. This is analogous to the versioning concepts. I will add some text that states this explicitly.
specification/2.0/README.md
Outdated
| |----|----------------|-----------------|-----------| | ||
| |`POSITION`|`"VEC3"`|`5126` (FLOAT)|XYZ vertex positions| | ||
| |`NORMAL`|`"VEC3"`|`5126` (FLOAT)|XYZ vertex normals| | ||
| |`TANGENT`|`"VEC4"`|`5126` (FLOAT)|XYZW vertex tangents where the w component is a signed float (-1, +1) indicating handedness of the tangent basis| |
There was a problem hiding this comment.
Sounds like a good candidate for 10-10-10-2 vertex format (ES 3.0+). Could be considered for the next glTF update.
specification/2.0/README.md
Outdated
| |`POSITION`|`"VEC3"`|`5126` (FLOAT)|XYZ vertex positions| | ||
| |`NORMAL`|`"VEC3"`|`5126` (FLOAT)|XYZ vertex normals| | ||
| |`TANGENT`|`"VEC4"`|`5126` (FLOAT)|XYZW vertex tangents where the w component is a signed float (-1, +1) indicating handedness of the tangent basis| | ||
| |`TEXCOORD_0`|`"VEC2"`|`5126` (FLOAT)|UV texture coordinates for the first set| |
There was a problem hiding this comment.
Normalized UNSIGNED_SHORT should be allowed here.
There was a problem hiding this comment.
I'll add normalized UNSIGNED_SHORT to all relevant attributes.
specification/2.0/README.md
Outdated
| |`NORMAL`|`"VEC3"`|`5126` (FLOAT)|XYZ vertex normals| | ||
| |`TANGENT`|`"VEC4"`|`5126` (FLOAT)|XYZW vertex tangents where the w component is a signed float (-1, +1) indicating handedness of the tangent basis| | ||
| |`TEXCOORD_0`|`"VEC2"`|`5126` (FLOAT)|UV texture coordinates for the first set| | ||
| |`TEXCOORD_1`|`"VEC2"`|`5126` (FLOAT)|UV texture coordinates for the second set| |
There was a problem hiding this comment.
Do we limit the number of UVs to 2?
If so, we should state this more explicitly somewhere.
There was a problem hiding this comment.
Yes. We need to limit them to something to prevent inconsistencies between implementations. 2 seemed reasonable to me considering we are running out of vertex attributes already for morph targets. I'll add more text on this.
|
+1 |
|
This addresses #812. |
No description provided.