Rob's Musings

So the Khronos group, the consortium which took control of the development of the OpenGL API (and some other related tools), recently released the specification for OpenGL 3.0. After giving the new spec some time to sink in, I've decided that my initial feelings of extreme frustration and disappointment were, indeed, well founded. I'll copy here most of the content of my post on the GameDev.net forums (where I go by the handle gfxnomad). This post basically responds to the question "Why is everyone so upset with the ARB/Khronos?" and secondarily "Why is the new OpenGL 3.0 spec so bad?"

First, the transition from GL 1.x to GL 2.x was a huge letdown. Then, they promised a clean, new, streamlined and generally more forward-looking API in Longs Peaks. Some people saw this as an exciting opportunity to revitalize the OpenGL community--to make it a bit more appealing to those on the fence, or those who have to do *nix/OSX 3D development. People were generally optimistic and expectant, and then they delay, an entire year, the release of the spec. Originally, the 3.0 spec was supposed to be released at Siggraph `07. The community was, to some extent, upset and disappointed with the delay. Finally, a year later, they release the OpenGL 3.0 spec and, low and behold, it's effectively a joke. Where is the removal of the old, unnecessary "cruft" (to use the operative term)? Where is the new object model? Where is the radically different design we were shown in the Longs Peak slides and newsletters? All of these things are completely absent, and in their stead, we are given the promotion of a number of extensions into the core and marking of some old functions as deprecated. The magnitude of this let down, despite what some might say, is hard to overstate. It may be true that nobody really expected OpenGL 3.0 to make OpenGL competitive with D3D from a "niceness of usage" perspective in one fell swoop. However, the modernization of the API would have placed it in a much better strategic position. Once OpenGL 3.0 was released, we were promised a series of fairly rapid, post longs peak updates; things like Mt. Evans. The hope was that the massive push to modernize OpenGL and the object model -- to streamline it and make it easier for both application and driver developers -- would place it in a position where rapid updates were possible, and it could be made competitive with D3D in an iterative fashion.

However, the OpenGL 3.0 that we were given pretty much crushes these hopes. There are two main reasons for this. First, since much of the major overhaul that was originally promised is nowhere to be found, all of the progress mentioned above will have to wait until the major overhaul actually does take place (if ever). This is a HUGE disadvantage. If Longs Peak was delivered at Siggraph `07, we'd probably be adding features and functionality to it at this Siggraph. If Longs Peak was delivered this year, then at least we could start getting our feet wet with the new API fairly soon. As it stands, we're stuck waiting again; in the mean time given OpenGL 2.2. The second reason the hopes are crushed is because Khronos has effectively lost the trust of the development community. The series of let downs that have been delivered means that many developers will not trust what they promise to deliver in the future. An active, healthy development community helps push the API forward, whereas a disgruntled, disenchanted development community leads to a stagnant and obsolete API in the long run.

Anyway, they have let us down before, but this is letdown on an entirely new scale. It's very disappointing, and frankly, a bit sad. This particular letdown may be the straw that broke the camel's back.