johnbue at msn.com
Wed Aug 1 16:32:11 New Zealand Standard Time 2001
Nathan F. Yospe writes:
> John Buehler <johnbue at msn.com> said:
>> Algorithmic generation of terrain produces perfectly wonderful
>> results. You just have to work on the algorithm. You don't stop
>> with fractal generation. You alter the algorithm in different
>> areas and at different fractal levels. You make the vegetation
>> type dependent on latitudes, altitudes, inclines and general
>> locations. The same with the wildlife. Run a simulator to
>> produce an expansionist empire that then collapses and leaves
>> ruins here and there. Run batch generators to produce individual
>> features that fit into the world terrain matrix, like caves,
>> ruins and such. To be honest, I haven't figured out an efficient
>> way to handle flowing water. I think it'll boil down to a batch
> Sorry for the second reply, but this has gotten me started again.
> There are those on this list who may remember me for exactly this
> kind of wild design... and even some who have accused me of
> blue-sky naivite. Those people might be interested to learn,
> should they still lurk, that I have secured myself a decent living
> pulling off this very kind of wild sceme, albeit in a slightly
> more practical industry, on a regular basis. When, and I do me
> when, not if, I return to the games industry, I will be with no
> qualms able to make these claims once more...
I find the idea that this would be a 'wild design'. What I was
describing was a little application that I mentioned here a couple
months ago. I got it to the point where I could see that it had a
lot more potential than just spitting out terrain with a variety of
features and just let it drop. The biggest lesson that I learned
from it is that if you're going to create a huge world, you better
have something to do while you travel around. There has to be
entertainment all around the player's character. That entertainment
might be as simple as fishing in a stream, skipping rocks or
climbing trees, but it has to be there.
I'll also note that folks here seem to be reconsidering the "it
hasn't been done so far, so it can't be done" philosophy.
> One of the tricks is having a two way replicable seed for
> generating the landscapes, affects and applied changes to the
> landscape, and predicting changes over a span of time reliably.
Can you go into more detail here? Are you suggesting that there is
a rough timeline for any given object to determine where it will be
or what condition it will assume at a given time in the future? If
so, that seems to be a somewhat limited treatment given that it
doesn't permit exceptions to that schedule. Unless the statement is
that exceptions are handled and that the gain is made when one
realizes that most objects do, in fact, follow their own schedule
without player-produced exceptions.
> Most of the random generation leads to an intense fractal tree
> with last minute instantiation of nodes, and no true
> leaves... which allows both a huge amount of generated/stored data
> mingling, and a reasonable means to avoid doing a bunch of work
> that will never be appreciated. Create that rose only if someone
> stops to smell it... until then, it's a featureless rose.
I favor this approach, but I can't tell if you agree with it or not.
MUD-Dev mailing list
MUD-Dev at kanga.nu
More information about the MUD-Dev