I guess the OpenGL vs Direct3D question has finally been answered. It's Direct3D unless you need cross platform compatibility.

Very sad.

Regards
elFarto

Although I can understand some of the frustration around here, I think you may be a bit too pessimistic there.

I've only skimmed over that specification for about 5 minutes, but so far it really doesn't look that grim.
Yes, there are no objects. Yes, there is still fixed functionality around. No, there has been no complete API rewrite. But hey, so what.
Look at the positive side. A good number of extensions with valuable features has been promoted to core, even some features that weren't initially announced at all. So, some stuff will be a bit less messy, which is good already.
Maybe "version 3.0" is a bit far fetched for what's in the spec, I'd rather have called it 2.2 or 2.3, but hey... what do numbers really mean. Remember when Netscape jumped from 4 to 6 in one go? :)

To be honest, this spec is a lot more than what I had expected to happen, and I see it as a very positive thing. In fact, I had expected that nothing happens at all, and I am very happy that we got at least half of what was promised, and a few goodies to comfort for the other half.
If you consider how many people/companies inside and outside of Khronos have excellent economic reasons against supporting and extending OpenGL, and how few people have a reason to support it at all, it's a miracle.

Hmmm, I must admit I was looking forward to a new streamlined OpenGL, but I don't find this new version that bad. You can see some of what they were aiming for in the new deprecation model. It does away with a lot of the cruft in OpenGL, including pretty much everything fixed-function. They've been quite brutal with what they've culled as listed in Appendix E.

I suspect this'll actually make life easier for driver writers as they'll write new drivers to support the new deprecation model with an OpenGL 2.1 emulator sitting on top of that. If the ARB is wise they'll provide the code for that emulation layer to help things along.

It's not the re-write they had promised us, but after that strange year of dead silence we kinda all knew something was up. But this just doesn't seem that bad to me. It's not a huge leap but it is a step in the right direction.

BTB, I've had the misfortune to create renderers for the Intel i915 in both GL and D3D, and their drivers stink on both. :-/

Sounds like lots of 'The sky falling' talk but I thought the sky had fallen years ago. I am using OGL/GLSL and it has everything I need. The trick is not to read the DX SDK :)

I don't think a clean slate would help drivers out since it would likely mean a new driver from scratch plus maintaining their old one. Honestly, do you think people would just run over to the oven for the freshly baked OGL 3.0 and say 'Finally, we can all stop sitting on our hands and start witting our OGL 3.0 driver'?

I learned roughly OS/2 years ago not to take sides. Just use OGL for what it's good for and leave it at that.

Wow, that is in fact a massive letdown.

Though, I interpret this as an artifact of the upcoming paradigm shift in graphics, introduced by Larrabee et al. ... Makes me think that the DirectX story is far from being told with Microsoft adding Compute Shaders to it either.

Part of the problem seems to be that there is no sufficiently good API to that future architecture yet, not even conceptually, let alone one that traditional things such as OpenGL/DX could easily be consistent with. So that no really future-proof way of evolving these is in sight atm.

[Edited by - pro_optimizer on August 11, 2008 1:47:17 PM]

Extremely unprofessional job by Kronos, they took an extra year to basically just add a couple of NV extensions as ARB extensions. If they were going to go back on their original intentions they should have made it clear much earlier. Don't even see Geometry shaders mentioned either so sub sm4 levels. I'm done with OGL asap.

Quote:Original post by Tesshu
I don't think a clean slate would help drivers out since it would likely mean a new driver from scratch plus maintaining their old one. Honestly, do you think people would just run over to the oven for the freshly baked OGL 3.0 and say 'Finally, we can all stop sitting on our hands and start witting our OGL 3.0 driver'?

Given that the original re-write from scratch plans were from ATI and NV I have to call that into question. Yes, they would still have to do some mantainence on the old drivers, however for the new they have a clean slate.

Right now when functionality is added into the core they have to check how it interacts with EVERYTHING down to glVertex() and the like. Stripping all this out would have given them a clean slate to start again and not have to worry about these old, outdated methods and how they interact with the newer functionality.

As for the 'deprecation model' someone mentioned; yes because deprecation has worked so well in other cases. People keep using it (see Java) even when there is a big 'do no use!' warning slapped on it, which means it has to stay in because as soon as you remove it it breaks something and BAM! loads of complaints.

The amusing thing, from what I recall there was always plans for a backwards compatible system, with contexts being shared between GL2.x and Longs Peaks so both could operate at once. Guess that wasn't enough...