Jump to content

Blender dropping COLLADA support


Masami Kuramoto
You are about to reply to a thread that has been inactive for 3209 days.

Please take a moment to consider if this thread is worth bumping.

Recommended Posts


Pussycat Catnap wrote:

Well that just kinda kills it for SL mesh now don't it?

Nothing's been decided yet, so nobody panic yet.  If worse comes to worst, and the Blender devs do cease official support for COLLADA,  I'm sure enterprising SL residents will just create their own COLLADA exporter for Blender, and release it to the public, just like they did for the excellent sculpty tools that so many Blender users now enjoy.

 

dontpanicoldT1.jpg

 


Pussycat Catnap wrote:

 

Why did SL go for collada and not something common like obj?

OBJ wouldn't be suitable.  Among other shortcomings, it doesn't support rigging. 

COLLADA was an excellent choice in that it is open source, it is currently supported by a very wide range of 3D modeling programs, and it's a very rich format with support for just about all the major scene components needed for effective transfer between programs.

Link to post
Share on other sites

I wonder if a python-only solution would be feasible, as it would allow the exporter to become an add-on and not need to deal with compile issues.

Although there's something to be said for using the OpenCollada library and being on the same page as all the other major packages right now.

Link to post
Share on other sites

I never understood the choice of Collada. It is an open source collaboration of some former game developers. To me it appears a very vague and under developed one looking at the Collada website too and reading wikis about it. OBJ is the professional standard of all 3D tools, and even SL uses it for objects and avatars as far as i could see. Strange that Collada choice was from my perspective.

Link to post
Share on other sites

Chosen Few wrote:

Nothing's been decided yet, so nobody panic yet.  If worse comes to worst, and the Blender devs do cease official support for COLLADA,  I'm sure enterprising SL residents will just create their own COLLADA exporter for Blender, and release it to the public, just like they did for the excellent sculpty tools that so many Blender users now enjoy.

Are you looking over my shoulder?

Oddly just before this issue came up, I'd just started digging into the Blender code to fix some Collada bugs. I do tend to agree that the current implemention in Blender is awkward and in the long term best replaced. So instead of fixing bugs, I've started on a Python implementation of a Collada exporter. And getting it working for SL is the first priority - it'll be free to all Primstar 2 and Avastar owners - I hope to have it ready before Collada support is dropped from Blender :)

 

Link to post
Share on other sites


Linda Brynner wrote:

I never understood the choice of Collada.

LL chose COLLADA because it's really the only format that makes good sense for what SL content creators need. They did not choose it lightly, or at random.  I'll explain as I respond to each of your other points.

 

 


Linda Brynner wrote:

It is an open source collaboration of some former game developers.

Sony is a former game developer?  Autodesk is a former game developer?

Linda, I'm not sure where you're getting your information, but COLLADA was first developed at Sony, in collaboration with other companies like Autodesk, Criterion, and Avid.  It has since been adopted by nearly every major 3D modeling platform, as well as a great many game engines, virtual world platforms, and other 3D applications.

COLLADA is now stewarded by the Khronos Group, a consortium of well over a hundred member companies, ranging from hardware makers like nVidia, ATI, and Intel, to software makers like Adobe, Sun, Google, and Mozilla, as well as game companies, telecoms, consumer electronics companies, large conglomerates, and universities.

 


Linda Brynner wrote:

To me it appears a very vague and under developed one looking at the Collada website too and reading wikis about it.

Terms like "underdeveloped" and "vague" aren't really applicable in this context.  Moving entire scenes between 3D platforms that each operate on entirely different, and often mutually incompatible, principles of organization is no easy task.  It used to require advanced specialized software, like PolyTrans, just to make it work at all.  In many cases, it still does.  So called "in-between formats" like COLLADA, FBX, etc., have been a godsend in this regard, for the past few years.

 


Linda Brynner wrote:

OBJ is the professional standard of all 3D tools

There is no singular professional standard, for anything in the 3D industry, in this regard.  OBJ is just one option among many, and not a particularly great one for any but the most simplistic of purposes.  It can only store the most meager of 3D model information, limited in scope to just elementary geometric components and basic materials, nothing more.

Just like with COLLADA, different programs write OBJ files in slightly different ways, leading to all kinds of inconsistencies and incompatibilities.  In cases in which all you need to do is describe a relatively simple model, with an easily discernible vertex winding order, and in which the materials assigned to that model are among the most commonly used, OBJ is usually (but not always) workable without any trouble.  When you get into anything more complicated or more specialized, however, then OBJ just isn't suitable.

 


Linda Brynner wrote:

even SL uses it for objects and avatars as far as i could see

What evidence do you have for this?

Granted, I'm not qualified to read the source code, in order to discern what kind of data is utilized internally to describe the geometry of the avatar and/or other models.  But I'd be very surprised if it were in fact arranged as OBJ data.  I have to assume that that would be pretty inefficient, programmatically.  If you know something I don't on this, please do share.

 

I do hope no one assumes that the mere fact that avatar mannequin models are available in OBJ format is at all indicative of anything that happens internally in SL.  Those mannequins are also available in Poser format, but I'm sure everyone realizes SL doesn't itself utilize Poser files in any way, shape, or form. 

Besides, if I remember correctly, the mannequin files were reverse engineered by a resident, and then given to LL.  At the time, LL had no plans of theitr own to do the work to make the models available, from what I understand.

 


Linda Brynner wrote:

Strange that Collada choice was from my perspective.

If you're aware of any other format that meets all the following criteria, I'd love to hear about it:

  • Open source
  • Supported by an extremely wide gamut of 3D modeling programs
  • Supports not just elementary geometric components and basic materials, but also rigging/bone weights/skinning.
  • Supports other features that SL does not yet utilize, but in likelihood will in the future, including advanced shaders, advanced materials, embedded animations, and physics, among many others.
  • Easily editable by hand, if need be, in any text editor

I can think of quite a few formats that each include some, or even most, of what's on that list.  I cannot think of any other at the moment that includes them all.

Link to post
Share on other sites


Domino Marama wrote:

Are you looking over my shoulder?

Oh, if only I could!  I'd love nothing better than to glean first hand insight into the genius that is Domino.  I mean that, most sincerely.  :)

 


Domino Marama wrote:

Oddly just before this issue came up, I'd just started digging into the Blender code to fix some Collada bugs. I do tend to agree that the current implemention in Blender is awkward and in the long term best replaced. So instead of fixing bugs, I've started on a Python implementation of a Collada exporter. And getting it working for SL is the first priority - it'll be free to all Primstar 2 and Avastar owners - I hope to have it ready before Collada support is dropped from Blender
:)

 

I figured you probably had something already in the works.  I suspected you wouldn't be able to stomach the existing implementation problems, and you'd be whipping up something great, even before I'd heard that official COLLADA support in Blender might be on its way out.  I'm thrilled to know my assumptions were correct.

Once again, Domino, you place yourself squarely in the top of my personal heroes list.  Don't ever stop!

Link to post
Share on other sites


Gaia Clary wrote:

.obj does not support weighting and armatures as far as i know.

Correct. 

Specifically, OBJ includes the following:

  • Vertex coordinates
  • Texture coordinates
  • Normals
  • Face definitions
  • Referenced materials (via associated .mtl file)

That's it.

Link to post
Share on other sites


Drongle McMahon wrote:

Hi Domino. Why can't the old illusoft exporter from 2.49b be adapted to work with the new Blender versions? Have the Blender internals changed too much?

It probably could. But I'm sure you can imagine the issues in converting ~400k of code to a different API.

It was also slow - it was written against a version of Python that didn't have elementtree,

The xml side would need changing from minidom to elementtree and the blender side needs changing to the new API. So the conversion actually becomes a total rewrite if you want to do it right. Even then, not all Collada features were supported.

Link to post
Share on other sites

Even the 2.49b version is buggy. If you work with it for some time you will find that blender occasionally hickups and terminates ungracefully with a crash. But restart and reexport works always (well, almost) at least for Second Life related content.

But honestly to all my knowledge the current Collada 2.59 or newer also works great for Second Life if used together with Masami's idea of rotating the rig bones towards the y axis. So its sad to hear that Collada might get kicked out as it just started to work for us.

On the other hand this move seems to vitalise developer brains a lot and i am very glad that the race has started :) There is plenty of oportunities to make it even better and maybe add some minor tweeks to make our life even easier. Just thinking loud about material indexes and order of objects ...

Link to post
Share on other sites

Yes. It's a long time since I exported anything without saving the Blender file first. A lesson well learned. 

When you are tired of thinking about ordering, try thinking about offline weight calculation. I think you'd need to get Falcon to let you see the right pieces of server code for the physics weight, but that would be a neat thing to have. A live weight display would be great for LOD optimisation.

 

 

Link to post
Share on other sites

So long as you and Domino are still hacking on primstar and avastar, Gaia, I suspect the utility of blender for SL will remain undiminished :)

The line I usually used when advising a blender newbie is that "You can do it without this addon, but you'll have a much lower level of frustration and keep more of your hair if you just go get this. A couple of better python coders than I'll ever be* have automated the gnarly bits of the workflow for you..."

 

 

 

 

* I'll give you a run for your money in perl but my python code is for crap

Link to post
Share on other sites

Reading through the email thread, it looks like the Blender folks will move Collada into a branch of the Blender developement. I think we would then see the Collada exporter/importer disappear from the basic Blender install.

Domino and Gaia have posts in the thread. It would be nice to see a Linden chime in too.

Posting to show support is probably not going to impress or move the Blender peeps. They have their Blender interests and we have our SL interests. I suspect for the Blender peeps keeping Collada in is a matter of, show us the code.

But, someone offering money or code is likely to have influence. I seem to remember a Linden saying something about the Lab working with Blender on Collada. If that is still the case, the Lab might fix the problem.

Also, Google uses Collada. So, it may be possible to get Collada into Google's open source coding projects...

Link to post
Share on other sites

It might be of equal interest to some of the technicians that about one hour ago OpenCollada has started to open their Generator for creating the Collada access libraries. I think that one major problem of the past was that this generator seems to have introduced some bugs. But it was not published, so nobody could work on it. And now as it becomes open Source, developers can start to address the problems right at the sources. Correct me if i am wrong...

So it looks very much like Blender has woken up the sleeping dragon. I was sure that this year will become exciting, although i did not expect it to happen like that :) Maybe Linden Labs can show some supporting initiative in this case. I believe that such a gesture would be very welcome.

Link to post
Share on other sites


Domino Marama wrote:

Chosen Few wrote:

Nothing's been decided yet, so nobody panic yet.  If worse comes to worst, and the Blender devs do cease official support for COLLADA,  I'm sure enterprising SL residents will just create their own COLLADA exporter for Blender, and release it to the public, just like they did for the excellent sculpty tools that so many Blender users now enjoy.

Are you looking over my shoulder?

Oddly just before this issue came up, I'd just started digging into the Blender code to fix some Collada bugs. I do tend to agree that the current implemention in Blender is awkward and in the long term best replaced. So instead of fixing bugs, I've started on a Python implementation of a Collada exporter. And getting it working for SL is the first priority - it'll be free to all Primstar 2 and Avastar owners - I hope to have it ready before Collada support is dropped from Blender
:)

 

This is really great news to hear. Thanks for your continued support for Blender. ^^

Link to post
Share on other sites

Archived

This topic is now archived and is closed to further replies.

×
×
  • Create New...