On Wednesday, October 31, 2001, at 06:57  PM, Jon Hylands wrote:

> I've emphasized this in all of my messages. Pushing pixels takes very
> little cost now, but scene management with today's complex levels and 
> high
> polygon counts is the big hit.
>
> OpenGL hardware doesn't help a single iota with scene management.

The last sentence is, of course, quite true -- most hardware support 
doesn't doesn't really hurt scene management.

Actually, unlike the inherently brute force processing inherent in pixel 
pushing, scene management admits highly complex and logic-deep 
algorithms.  I disagree that the "big hit" is manifest in scene 
management as suggested, or that modern state-of-the-art games are even 
close to as processor-intensive as those of past years.

I had this out with John Romero a few years ago, when the technology 
mismatch for pixel pushing was not even close to the range today; at a 
time I took your position.  He was trying to convince me that 
first-person shooter medium is destined to come beyond the 
processor-deep no-brain no-game doldrums in which it was buried; 
pointing out the particular technical changes that had already 
occurred.  Having seen the insides of these games, I see now that he was 
right and I (then) and you (now) are wrong.

Game programming companies, the particular ones you described in earlier 
posts in fact, are desperately seeking to evolve from the graphics-deep 
technical skill pushers into a content-deep game design plant.  They 
realize that the time for game mechanics (in the auto mechanic sense) is 
the past and game mechanics (in the game design sense) is the future.

Jon's argument ignores reality -- and more significantly the particular 
lesson that Alan and crew taught us more than twenty years ago.  When 
bitblt technology was considered deep dark magic, the idea that it could 
be delivered in a full GUI interface from a uber-slow, memory-managed 
byte-coded p-machine was ludicrous.  But they did it, and on machines 
many orders of magnitude slower than a Palm hand-held.  The model of 
identifying complex hunks of technology as data-driven primitives and 
executing them is enormously powerful.

Jon's nihilistic suggestion that Squeak is incapable of doing ubergames 
is, IMHO, naive.