[Opensim-dev] How to Implement a wind simulation

[Dahlia Trimble]

I'm interested in possibly adding support for a wind simulation in opensim,
and I've been doing a little research.

According to http://wiki.secondlife.com/wiki/LayerData there is wind data
sent for 16x16 meter cells in a region. I'm not sure yet how often it is
sent, but libsecondlife has a data throttle for it so I'll assume it's
periodically updated. I have tested wind-responsive flexy prims on the Agni
grid and they appear to move with strikingly close synchronization between
different viewers. They do not correspond well to wind vectors obtained with
llWind() in the same 16x16 meter section of a region, there seems to be a
several second time lag *as if* the llWind() data have a much lower update
frequency, or a time lag on the order of many seconds.
http://wiki.secondlife.com/wiki/LlCloud mentions a relationship between
clouds and wind but is lacking detail. These data are encoded with a custom
discreet cosine transform routine similar to jpeg before encoding into
packets to be sent to the viewer.

Given the 16x16 meter cell size, I think a fairly simple model will suffice
for the purpose of adding realism to the user experience when using the
Linden viewer. Input requirements may be as simple as region time of day and
sun position, and perhaps water level and average terrain height for a cell.
Such a model may need little CPU and memory requirements and shouldn't
noticibly impact sim performance. More realistic simulation may require a
higher cell density and prim mesh data from the physics engine and could be
offloaded to a seperate thread or external process if necessary. Output from
the model should be sent to the viewer and made available to the scripting
engine to allow access via llWind() and llCloud().

I'd prefer to initially develop a simple model this with a plug-in
architecture that would provide the necessary hooks so more elaborate models
could be developed at a later date for those who may have interest in such
an endeavour. I realize that there are some concerns with the current
architecture and possible alternatives are being discussed, so I would ask
those to consider these needs in their discussions. I originally posted some
ideas about this topic to that thread but they appear to have been buried in
the chatter so I'm posting this as a new topic with hope that it may gain
more exposure.

Please comment!

Thanks,
-"dahlia"
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://opensimulator.org/pipermail/opensim-dev/attachments/20080625/15f91ba1/attachment-0001.html>