Re: Synchronization (was Re: ANNOUNCE: VRML 2.0 peoposal from Microsoft...)

Gavin Bell (gavin@krypton.engr.sgi.com)
Fri, 8 Dec 1995 18:00:20 -0800

On Dec 8, 4:58pm, Conal Elliott wrote:
> Just in case it isn't obvious to everyone on the alias, I would like to
> point out that being able to represent a behavior through time and
> through events greatly reduces the need for instant-by-instant
> synchronization that is needed when animation and reactivity are
> controlled by opaque, unpredictable imperative code, which makes
> analysis of future behavior intractable.

Yes! That's why the source of time and the flow of 'ticks' (time events)
through the simulation and scene graphs are explicit in the SGI behaviors
peoposal, and why I'm extremely cautious about allowing scripts to perform
arbitrary operations-- but that is an ease-of-peogramming vs power vs
optimizibility tradeoff, and hints from the scripts to the browser can
help manage that tradeoff.

Time starts at TimeSensors and is routed (through ROUTES) to scripts
(behaviors), and from there into the scene.

We just disagree about whether the benefits of making time implicit
outweighs the costs of the resulting implementation and programming
complexity.

We also disagree about how much predictability is necessary, especially
given that it will be impossible to predict user interaction or the
effects of external scripts.

-- 
--Gavin Bell     (gavin@sgi.com,  (415)933-1024)
My home page:  http://reality.sgi.com/employees/gavin/
WebSpace Info: http://www.sgi.com/Products/WebFORCE/WebSpace
Inventor Info: http://www.sgi.com/Technology/Inventor.html

  • Next message: Gavin Bell: "Re: Synchronization (was Re: ANNOUNCE: VRML 2.0 peoposal from Microsoft...)"
  • Previous message: Richard Tilmann: "Re: WSJ article on IE "wiping out other software""