back
11 comments
Having skimmed the book briefly, I find it oddly selective and incomplete in what it chooses to acknowledge and what it chooses to leave out.

Examples:

* In the global illumination section, photon mapping is given a whole subsection, but there's no mention of Veach whatsoever, despite the fact that photon mapping is rarely used in practice today whereas Veach is the foundation of all modern path tracing. This on the level of like, writing a History of AI book that doesn't find it important to mention Frank Rosenblatt/Yann LeCun/Brendan Frey's foundational contributions to deep learning, or writing a History of Operating Systems book that doesn't find it important to mention the Mach project, or writing a History of Programming Languages book that chooses to skip mentioning lambda calculus.

* Speaking of photon mapping; the book provides references or names for many techniques, but then spends a whole section on photon mapping without bothering to name who invented it (Henrik Wann Jensen) or where it was invented (Technical University of Denmark).

* Stanford isn't mentioned at all in either of the chapters on early graphics research labs, even though from the late 80s through to the early 2000s Stanford was one of the biggest powerhouse labs in rendering, simulation, and vision research.

* In the section on Physically-based modeling, Witkin & Baraff are mentioned extensively, but Ron Fedkiw doesn't get a mention anywhere in the book despite the fact that PhysBAM is likely the most influential simulation package ever created.

* In the chapter on animation software, extensive information is given on Maya and Houdini, but there's no mention whatsoever of large popular packages that _predate_ Maya and Houdini, such as Blender, Cinema 4D, 3ds Max, etc.

* From this book you'd think that ILM and R&H were the only companies to make any significant pioneering steps in using CG for VFX; never mind the huge contributions starting in the early 90s made by the likes of Digital Domain, Weta, MPC, Framestore, Sony Pictures Imageworks, and others.

* The book contains straight up hoaxes, like this debunked image: https://www.snopes.com/fact-check/1954-home-computer/ [edit: ah, I see kmoser noticed this too and mentioned it in another comment on this submission]

Speaking as someone who works in the CG field, there are many many more obvious, bizarre omissions.

Yeah, a lot seems kind of weird. Many things get mentioned in odd places or credit seems somewhat misplaced.

* It mentions James Kajiya, but David Immel usually also gets co-credited for the rendering equation. (He's credited in the glossary, but not in the main section on global illumination!)

* Particle systems mentioned elsewhere, but not in the ILM (Lucasfilm) section.

* "1968 also saw the arrival of Ray Tracing developed by Bell Labs and Cornell University." Here, it seems to be confusing Appel, who published on ray tracing as a visibility algorithm in 1968, with Whitted who published on recursive ray tracing as a shading algorithm in 1980 while at Bell Labs. One needs to be really careful about conflating ray tracing as a visibility algorithm vs as a shading algorithm.

* In the timeline: 1987, "Willow (Lucasfilm) popularizes morphing" vs. 1988, "Willow (Lucasfilm) uses morphing in a feature film".

* "One approach to taking more responsibility from the CPU was done by the GeForce 256 from NVIDIA (the first graphics processing unit, or GPU)." Next paragraph: "3dfx introduced their GPU, the Voodoo in 1996." Umm, the GeForce 256 came out in 1999; 3dfx was first here, though I believe that NVIDIA was first to coin the term GPU as a marketing term. And of course there were things like the SGI Onyx with its 3D accelerators before either (and SGI was discussed in the previous section).

* Shading languages? BRDFs? Shadow mapping? Free form deformation, skinning, and non-rigid animation? Dithering and halftoning? Digital color? POV-Ray?

There should be a clean split between fixed function accelerations, like 3DFX did, and when you could compile simple code to be run on the GPU. (Shaders?)

Also, at some point (NVIDIA?) you could start downloading graphics (or computing results) back to the CPU.

Yes, definitely. IIRC, the fixed-function aspect is where NVIDIA's marketing made the distinction between their "GPU"s and the graphics accelerators that had gone before. That sort of thing would have tied in nicely with a discussion of the evolution of shading languages if this book had had such.

And credit where it's due: I believe that NVIDIA also led the way on going from abusing the graphics pipeline to do GPGPU compute workloads to opening it up to much cleaner approaches like CUDA.

Does the book even mention B-splines or Nurbs? Subdivision?
Yes, those are touched upon.
I didn’t notice in my glance through… but in general yeah, it’s a really weird survey
Looks like their fact-checkers are asleep at the switch. This page https://ohiostate.pressbooks.pub/graphicshistory/chapter/2-2... contains a photo of a well-known hoax https://www.snopes.com/fact-check/1954-home-computer/
I found small factual errors in the first 2 pages I skimmed, e.g. the clock speed of the Amiga 1000’s CPU and the names of its custom chips. Hardly material errors, but the apparent lack of review by anyone who knows the subject matter would make me wonder what else was incorrect, were I to read the rest of the book.
Seeing a Gunstar on the cover gave me a warm and fuzzy feeling :)
Great resource. Thanks!