gt0852c%prism@gatech.edu (Daniel Rothman) (06/08/90)
Two perhaps controversial opinions:
1) someone stated earlier that a standard VR development space should
be MS-DOS compatable. I disagree. The primary "bright-eyed larvae"
are going to be college students. The main op-sys on campuses
today is rapidly becoming/is *nix. I think this is a good place to
start.
2) it seems to me that the most plausible way to go about building
a GOOD VR is going to be in an extremely modular fashion. The best
development tools we have at present for truly modular programming
are object-oriented ones. I believe that if we could get a large
base of programmers to agree on a fundamental idea of object-oriented
VR then we'd have a much easier time of it.. Better yet, lets all
agree to use an industry standard that's also realitively cheap, like
C++.
Although this sounds prejudiced and narrow, there are good reasons
for it. a) Object-oriented stuff allows of a higher level of
abstraction for the programmer. This may be of paramount importance
in such an information-rich environment as C-space. b) It allows a
much higher degree of portability - as long as all the objects of
the same type receive and answer in the same manner, there is no
problem with interface. This will be especially helpful for such
operations as "show polyhedron".. c) it has been shown that for
very large, complex systems (such as C-space) dynamincally binding
languages sometimes run FASTER, despite messaging overhead. (I
believe this is because the need to do some object-like things in a
standard language creates massive overheads).
So let's get some debate going on here over STANDARDS. We've had a
lot of ideas, most of them very sound. Let's walk the walk, guys.
uxmal - be nice, i'm an endangered bright-eyed larvae :)