[sldev] A thought on how to lower resources used by sensors

Kelly kelly at lindenlab.com
Mon May 18 13:03:35 PDT 2009


Henri Beauchamp wrote:
> On Mon, 18 May 2009 13:16:29 +0200, Ambrosia wrote:
>
>   
>> On Mon, May 18, 2009 at 04:16, Nexii Malthus <nexiim at googlemail.com> wrote:
>>     
>>> I think rather than keeping the problem on the LSL side, it would be more
>>> suitable for being a viewer-side feature entirely that replaced the
>>> necessity of such silly HUDs and Gadgets. Finally removing the source of the
>>> problem alltogether, and virally so due to the big advantages of being able
>>> to be tied into the user interface.
>>>
>>> - Nexii Malthus
>>>       
>> Which would not remove, at all, the problem of having to use sensors
>> in an object in order for the object to detect who is nearby. It would
>> be much more feasable to have a function return a list of avatar keys
>> in the sim, and to call this function once whenever the agent count in
>> the region changes.
>>     
>
> I think the best solution would be to allow sim-wide sensors
> (llRegionSensor() ?)... This would remove the need for probes, just like
> llRegionSay() suppressed the need for llShout() relays in the sims...
>
> Henri.
>   
Sensor events are kind of weird.  They were designed specifically to 
give a real-world quality to gathered data.  If you have a sensor in the 
real world, the data it can pick up is generally directly related to 
where the object is and some range that it can sense information, and 
probably a direction.  There are definitely uses for this kind of 
information gathering in Second Life, "what objects are near me" and the 
various permutations thereof are definitely useful.

However they do have problems.  They are limited in what data they 
collect since they scan and store a snapshot of the data with the script 
- we can't really extend how many results or what data is gathered just 
because of the memory footprint involved.  The legacy script formats add 
additional difficulties that are less insurmountable but still trouble.  
By their nature they are at least a little more load intensive than just 
accessing data the region already has in memory would be.  If you 
extrapolate from this just a little bit you can think about why region 
wide sensors wouldn't be that great.  We could still only return the 16 
results and the existing types of data.

As has already been suggested in this thread I believe, we are much more 
likely to investigate non-sensor calls that directly access information 
the region already has in memory and don't try to mimic real world 
behaviors or the sensor pattern.  llGetAgentCount() or 
llGetAgentsInParcel for example, that could return data directly.  I'm 
not saying we are working on these or even have plans for them, and 
there may be privacy issues that come into play for some of these types 
of calls.  I'm just saying that I think this is more likely to be the 
direction we head than extending or mimicking the llSensor pattern.

 - Kelly
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.secondlife.com/pipermail/sldev/attachments/20090518/29729d65/attachment.htm 


More information about the SLDev mailing list