I'm not digging the complicated "JFK" theory here.
Originally Posted by zonyl
Keep in mind that Rubinstein left Apple long before even the first iPhone came out, and the 3GS hardware specs would not have been known at the time Palm hired away any Apple people to work on the Pre. Palm chose the TI OMAP 3430 platform, which allows them to avoid a ton of hardware engineering -- TI takes care of the bulk of it. Starting from this point, Palm can basically work on the device's industrial design and then get everything to fit in the case. But the fundamental system architecture was done by TI, not Palm (which is a big difference from how Apple does it -- they lay out their own architecture lines, and the ARM-based processor in the 3GS even has an Apple logo on it). The 3430 happens to contain a GPU, whether or not the client (in this case Palm) wants to use it.
The distinction between 2D and 3D is not as important as the distinction between software (pixel) rendering and OpenGL rendering. As you might know, OpenGL has major benefits in terms of graphics acceleration regardless of the "dimensions" of the scene. In fact, Apple uses the GPU to do massive amounts of texture rendering for their 2D interface, which is why many aspects of the iPhone (and Mac OS X) are so smooth and fast. Even trivial 2D content is accelerated.
I think Palm made the conscious decision not to take advantage of OpenGL. This was likely driven by their desire to go with a web-app based SDK for their apps and third party apps. Until webGL debuts middle of 2010, there is really no way to fully use OpenGL in a web app. So I think the choice to use the web app approach is what drove the choice to *not use* the GPU (because they couldn't). It's as simple as that. If the GPU and graphics acceleration was more important to them, they would have been driven towards a native SDK.
So really, we are where we are now because Palm took the web app approach and there was no technical way to use the GPU from that approach. It's as simple as that, and you don't have to invent complex theories. They could not have used the GPU with the current approach at the time the device debuted even if they wanted to. As we move forward towards mid 2010, that will likely change.