Siggraph, wednesday 30

[ton]

The yearly OpenGL BOF usually is a presentation from ARB (architecture review board) members to present the advances of last year, and next year's plans. To my surprise they meeting didnt mention the long awaited opengl2 launch, but restricted to presentations on the shading language, new buffer extensions, OpenGL ES (embedded) and java-opengl. We might have missed the big news, because we were five minutes late, so after the presentations I went for a chat with a few of the ARB members to get my answers.

- OpenGL 2 specs now have been frozen, but the 'conformance' tests have to be agreed on yet
- expected first official launch not before siggraph next year
- even with OpenGL2 we'll have to live with extensions and versions... a standard like OpenGL is supposed to be, just cannot be enforced on (nor frozen) because of specialiasts needs from individual companies
- OpenGL 1.5, which should be available now, basically has the features 2.0 will have, but with a different naming for functions (not ARB_ extensions anymore)
- if we want to move on with OpenGL2 for a Blender3, we can also choose to strictly limit support to gfx cards that officially conform to ARB standards, meaning limiting choices to ATI, 3DLabs, Sun, SGI, and probably nvidia...

Nvidia was obviously NOT present at the BOF, which was - I think - representative for the behind the scenes political fight over the Cg vs.OpenGL SL standard. Too bad... also the role Microsoft wants to play in OpenGL is unclear... they wont 'help' but according to the ARB members, they wont 'attack' either. My question on the patent issue (MS has shader patents that might be enforced on) didnt give a clear answer either, other than "OpenGL has potentially 100s of patent infringing technologies implemented, non of the ARB partners ever was being sued over it, it's just unlikely anyone will". ARB partnerrs, still a strong part of the industry, will likely enforce their own patents, if attacked... patents being used here as a firewall.

In general, I've still got a very positive feeling of moving to OpenGL2 in the future. The specs are great, and creating support for this in a new Blender3 tool will give exciting possibilities. It just all takes so long.... with Direct3D becoming more and more the industry standard for consumer based 3d (games). Probably OpenGL will keep playing a (restricted but relevant) role in the 3d workstation & creation market. Which is fine.

-Ton-

[xype]

I don't think DirectX really has much of an advantage over OpenGL - at least with OpenGL each vendor can release nice extensions for it's special needs, dunno how possible that would be with DirectX.

Also Blender 3 will probably take about one year to start coming to life anyway and in the meantime OpenGL 1.5 could be used as a platform for it's development.

As for the patents - it's standard practice in the whole IT industry. Each company has as big a patent portfolio to prevent other companies to sue them (the only company trying to look like being able to sue is SCO, but SCO is crazy). This is good for the companies, since they have an ace in the sleeve in case things get nasty, but bad for new competition since a graphics card manufacturer can not to enter the market without buying into an existing company. If anything resembling real competition would come up ATI an nVidia could use their patents to sue them out the market (unless the new player comes up with a really unique, non-patented approach to 3D graphics). Microsoft basically bought SGI's patents in order to be able to play the game in the console market, imho.

[eskil]

Hi everyone.

Im not at Siggraph this year, and it saddens me. I try to tell my self that i it doesn't matter because Siggraph is smaller, there is less new things in the pipe, and i have been to most of the courses, but it doesn't really work...

1.5 is the "test" release of 2.0, it contains the same things but this is sort of the test run, so start coding!

I have been out traveling and have not been home much, but im working on verse killing that last bugs. This is a lot less fun then writing new apps so i have started rewriting connector and it will be out soon.

I have also started porting Loq Airou and its a lot of work, verse used to use integers but now use doubles for all vertexes so there is a lot of work on an app that uses that much geometry manipulation.

The IRC is starting to get active too.

E

http://www.blender.org/modules/verse

[Money_YaY!]

Why not just plan Blender 4 for OpenGL 2.0 .
And get 3 over with before the next year. Alot still needs to be odne ,yes, but still a 3.0 release would spark some more intrest in the indy.

^o^ whatever I just want a "real" unwrapping and game engine.

[dreamerv3]

From the way the meeting went, it seems that the IHV's ala... 3DLabs, ATI want to "test drive" the New ARB extensions in 1.5 with Nvidia basically choosing to market it's Cg variant, to the high end vfx vendors as a shader toolkit, (they still don't have advanced viewport rendering and Mayas' "Hardware accelerated rendering" is nowhere near the potential these cards present, the api is sitting right there, but "Lets use Cg" "Look at my pretty node based click and drag shader tool! Did I mention Nvidia supports it on Maya?"

To be fair Nvidia reps were saying that Cg was very similar to HLSL BUT guess what, if they want to be on my list of future card vendors they better fully support the GL standard including all the ARB extensions. Do I see a blender HCL on the horizon?

I can't belive these people, there's no innovation going on.

I had a demo rep do a "Hardware accelerated render" for me it was a flying logo frame est, 15,000 polygons a few lights (no volume) a simple fire particle emission with some swirl, the "Hardware based render" took about 10 seconds for that one frame of 640x480"

Then I said "Now could you please simply render the exact same scene via software"

Same scene took about 12 seconds.

Me: "Um I'm sorry but based on the hardware processing power you have here, and the amount of multipass being dumped on this card there's no reason I shouldn't be seeing 1 fps" If you're wondering why I etimate 1 fps for a 15k poly scene it's because, there WERE some nice specular and shading effects.

The maya rep:"But! wait a second" He said

"It might seem unimpressive with lower polygon scenes and simple particle systems, but when you scale it to 400,000 polygon scenes with lots of effects you see a dramatic increase in rendering speed"

Me: "So what you're sying it that I can't have a system where I simply interact with the application and never press render, perhaps only scale the detail levels up and down for interactivity with perhaps a (visual complexity) value or slider, but essentially never press render, to have the hardware simply render the scene with shadows and bump maps and cubic reflections and to then get in position in realtime, scale up the bump and shadow and reflection accuracy and hit save in that position, maybe even have a final render quality setting and a interactive quality setting which can be configured."

The maya rep: "No we can't do that" (followed by his attempt at explaining why he had no idea why anyone would want to have such a capability"

So when I say there's no innovation thats what I mean, the most innovative thing I saw was the hash booth, they had "auto rigging" where you draw the skeleton and then select "auto assign bones" and the funny thing was that it actually worked.

As far as OpenGL 2 is concerned, the extensions approach while not ideal is not a bad one, what can happen now it GL 1.3 support planned into the Blender 3 spec, and the rendering subsystem can actually be abracted from the GL layer via an extension mapping layer, this would allow an easily configurable way to adjust to the volatility inherent in this process, you could customize blender's GL support directly within blender adjusting directly to the application in question based on what your hardware supported.

The down side is that an opengl 1.2 card would not let you create very good looking output, and I'm not too keen on dropping functionality, the good news in that mesa can actually pick up the extensions that in the most unlikeliest of circumctances a certain board wouldn't support and let blender operate.

I hate this BS.

More static from M$... (oh we MIGHT have IP patents embedded so far up your a$$ even we don't know where they are but we are choosing not to attack you)

As far as I know ARB_ extensions are supposed to be standard GL extensions and as such should be supported by all vendors, and for what we need essentially in 1.5 the hlsl has been implemented as arb extensions.

We an just work based on what is a standard opengl extension and advance as the standard GL extension body grows and solidifies.

The likely outcome for GL 2 is 12 months both ATI and 3DLabs want to test drive the new extensions which means they could or could not change as far as naming and some other minor details.

I think I like this approach better than the GL remapping layer I thought of before.
Why do more work than you have to and why chop blender up into "functionality modes"
Thats too uncomfortable for me.

The GL 2 standard extensions might be preliminarily agreed upon as early as 03/04 as the slide @ the BOF indicated, but a "pure" GL 2 with much of this stuff in the core would come later on "perhaps in 2005 as I heard one of the speakers mention at question time with a low voice too" I htink he was one of the ARB speakers.

Well lots more can be done in other places before then, tomorrow I'll be working on my map for the location of functionality in the .c files.

"Run Ton Run!"

Later for now!

[thorax]

- if we want to move on with OpenGL2 for a Blender3, we can also choose to strictly limit support to gfx cards that officially conform to ARB standards, meaning limiting choices to ATI, 3DLabs, Sun, SGI, and probably nvidia...

Or abstract the feature set away so that OpenGL libraries can be interchanged.. But it would require intense thought about what
basically blender needs and what additives people would want..
Not sure how hard it would be to abstract away from the libraries..

This may be a leap of imagination, but imagine if OpenGL was like a plugin, by use of some glue code.. So internally blender uses the original OpenGL, but if the plugin exists it uses card specific directives..
But make it so that the objects in blender can have any amount of
specific detail, but levels of detail like for basic OpenGL no graphics card the object is red, for medium case its OpenGL with graphics card that supports bitmaps, and the ball is red and rocky looking but no bump maps.. And for the high detail the ball uses custom procedurals to
define the bumpiness.. The idea being that no matter what version of blender you use, you can still use blender.. It might make the blend files hug though..

Would it be possible to do this without a close coupling of the libraries with
blender's source? Particularly without having multiple #ifdef throughout
the source code? Some how make it a runtime problem and not a compile time problem? Then someone could develop a OpenGL2 emulator
for paralization across a super computer for instance, without custom hacking the blender code and recompiling? It would be the difference between having to debug 100 spinoffs of blender using different
OpenGL source and debugging a single source and the gluecode/libraries
seperate..

Nvidia was obviously NOT present at the BOF, which was - I think - representative for the behind the scenes political fight over the Cg vs.OpenGL SL standard.

Well NVidia is losing out in "Maximum PC" magazine, ATI is out ahead on
graphics runtimes for games.. MaximumPC says their demos tend to use special features of the card to come out ahead of ATI but ATI is more
precise on what it does.. Their favorite thing to pick fun at is the NVidia
cards with the huge cooling fan attached.. If NVidia doesn't play along
they could go the way of Voodoo..

Too bad... also the role Microsoft wants to play in OpenGL is unclear...

Typical of Microsoft.. Keep the crowd mystified.. I wonder how
good Intel and Microsoft's relation is, Intel is fixing to develop
chips with about four times the resolution of the current process..

they wont 'help' but according to the ARB members, they wont 'attack' either. My question on the patent issue (MS has shader patents that might be enforced on) didnt give a clear answer either, other than "OpenGL has potentially 100s of patent infringing technologies implemented, non of the ARB partners ever was being sued over it, it's just unlikely anyone will". ARB partnerrs, still a strong part of the industry, will likely enforce their own patents, if attacked... patents being used here as a firewall.

There is a term called patent flooding, where a competing technology
patents everything surrounding a competitors technology so they can't move out of the confines of the specific technologies.. I had a friend in school who wrote software that would generate 3D plots of closeness of
patents and said he could literally see the process at work..

Time to find out if the US patent office can point out that the patents
make use of hardware technology.. I think software patents are lame,
especially if they can be reduced to math.. To make something patentable you have to have some kind fo hardware dependency, a proof that the software is a part of a hardware process. I'm wondering what the patents are and if they are enforceable.. I guess they can implement software in hardware, then can patent the hardware.. But you should still be able to
make the software..

In general, I've still got a very positive feeling of moving to OpenGL2 in the future. The specs are great, and creating support for this in a new Blender3 tool will give exciting possibilities. It just all takes so long.... with Direct3D becoming more and more the industry standard for consumer based 3d (games). Probably OpenGL will keep playing a (restricted but relevant) role in the 3d workstation & creation market. Which is fine.

-Ton-

I'm sure the games market will eventually want to get away from DirectX
because Microsoft will use it to leverage themselves in there, then purchase Activision and other companies, and soon we will be buying all our games from Microsoft.. That's how they work, they get you dependent on them, then they turn around and screw your market and take over..

[thorax]

If anything resembling real competition would come up ATI an nVidia could use their patents to sue them out the market (unless the new player comes up with a really unique, non-patented approach to 3D graphics). Microsoft basically bought SGI's patents in order to be able to play the game in the console market, imho.

You don't get anything by sueing just loss of money to lawyers, its used
defensively as a threat of sueing as its costly.. Anyhow, the best way that
they could compete with Microsoft is open sourcing something that ruins some of Microsoft's markets.. I wonder if this is possible, maybe say an investor of NVidia or ATI is rather large and is competing with Microsoft in another area, they could turn around and do something drastic..
I wonder who is going to be the sacrifical lamb here..

[Michel]

Why not just plan Blender 4 for OpenGL 2.0 .
And get 3 over with before the next year. Alot still needs to be odne ,yes, but still a 3.0 release would spark some more intrest in the indy.

^o^ whatever I just want a "real" unwrapping and game engine.

Hi,

if you just want a blender 3.0, I can pump up the version number just for you. Of course, no added functionality will be available, no new internal structure, no new concepts, nothing. But, you have a blender 3.0 version

Sigh, what's it with people and major version numbers? We're not Microsoft!

With regards,
Michel

[xype]

I wonder if this is possible, maybe say an investor of NVidia or ATI is rather large and is competing with Microsoft in another area, they could turn around and do something drastic..
I wonder who is going to be the sacrifical lamb here..

Noone will be a secrificial lamb and most certainly not any company that has a chance to ship their GPU in the next revision of XBox, that's for sure. Any why the heck even attack MS, it would be stupid right now.

[Money_YaY!]

Why not just plan Blender 4 for OpenGL 2.0 .
And get 3 over with before the next year. Alot still needs to be odne ,yes, but still a 3.0 release would spark some more intrest in the indy.

^o^ whatever I just want a "real" unwrapping and game engine.

Hi,

if you just want a blender 3.0, I can pump up the version number just for you. Of course, no added functionality will be available, no new internal structure, no new concepts, nothing. But, you have a blender 3.0 version

Sigh, what's it with people and major version numbers? We're not Microsoft!

With regards,
Michel

Ha! That is just the point. People and the indy just love solid point versions.
You could put the ODE game engine in it as the Commiters are working on, the plugin script stuff, and theimport export stuff. That would be enough to be called a MAJOR upgrade. Oh! and the theme gui stuff.

There a new gui. Export, just like blender needed years ago, GAmes again, and Super powered python !

Part 3 !

^v^ then next year do all of that other stuff for 4

more intrest=more coders

[dreamerv3]

Excuse me?

Just pump up the version number add in some half baked features and undigested concepts and call it a release?

What do you think this is commercial software development?

[JA-forreal]

Excuse me?

Just pump up the version number add in some half baked features and undigested concepts and call it a release?

What do you think this is commercial software development?

Aren't there enough 3d software options for people like this, maya , max, etc? I sometimes wonder why they even bother to post.

I don't know about the rest of you Blender members but Blender suits me just fine. And great news from Ton of future Blender versions makes my Blender experience even better.

Would a DirectX version of Blender be possible? I don't see this a bad thing for us windows Blender users.

[xype]

Would a DirectX version of Blender be possible? I don't see this a bad thing for us windows Blender users.

And I don't see it making any sense, at least not until the game engine gets an overhaul. Right now OpenGL can be developed by many users on many platforms, a DirectX port could last a few versions, but what if the original author has no time for it after that? No Linux/Mac/BSD developer could help out, as they could with OpenGL stuffies.

And Blender uses OpenGL for it's interface, not sure you could mix that with DirectX, unless for the player executable only.

[Money_YaY!]

Excuse me?

Just pump up the version number add in some half baked features and undigested concepts and call it a release?

What do you think this is commercial software development?

Yeah. just realy good free software.

I am just giving input. There are like only three or four OSX develoers right now. That really needs to go so up ! And all of the coders that make games with Crystral Space and others could boost the code flow up sooooo much over here. But they still seem to have no intrest.

So mat, just time.

^v^

[thorax]

I wonder if this is possible, maybe say an investor of NVidia or ATI is rather large and is competing with Microsoft in another area, they could turn around and do something drastic..
I wonder who is going to be the sacrifical lamb here..

Noone will be a secrificial lamb and most certainly not any company that has a chance to ship their GPU in the next revision of XBox, that's for sure. Any why the heck even attack MS, it would be stupid right now.

I wasn't talking about the graphics world strictly or games.. I
was talking more about Microsoft and all the various fields where they apply the same power monging tactics.. Eventually they will have stepped on enough people..

And to the other guys about the revision numbers.. I think revision numbers was something that came in with commercial sales of software, I think its just there as an indicator of progress.. It would be stupid just to release something with a false indicator.. However the other approach is to release versions with letters attached.. I guess the other way out is to rename the product altogether can call it version 1.0 .. Maybe that would
get us a whole new crowd of users.. I know lets just make blender look like something new and call it something else..

I know we could call it "Somethingelse 1.0"
and have ads with some handsome guy and girl at a bar, and
the guy saying "I got this new ad done at work, I want you to have a look.. Check this out".. The Girl says "gee that's really hot, what did you use?".. He says (with a gleem in his eye) "Its is a secret, but I will tell you its Somethingelse"..

Okay that's got to be better than anything I've seen on TV recently.. Apart from the frankenstien selling arthritis medicine..