Jump to content

Profaitchikenz Haiku

Resident
  • Posts

    2,835
  • Joined

  • Last visited

Posts posted by Profaitchikenz Haiku

  1. If you are moving the pose objects using llSetPos or SetRegionPos in increments using a timer you can expect some jerkiness when avatar presence increases, because the timers will be slowed or paused according to loading.

    If you are using KeyFramedmotion then you will be less affected by avatar loading, with one exception:

    If the KFM command is issued at the exact time that an avatar with scripts enters the region and causes one of the momentary freezes, the KFM might be delayed. Similarly, if you are expecting a moving_end event and it happens to coincide with an avatar-arrival spike, the moving end will be delayed. Once the KFM command has been issued and is in hands of the server,  movement seems far less affected by avatar presence than I had been expecting. It is the beginning and the end that shows the most susceptibility.

  2. On 3/29/2022 at 9:44 PM, Hooten Haller said:

    Is there a way to make a region be heavily loaded without trying to find 60 accounts to log in?

    In my (unpublished) exploration of how to measure a region's performance using scripts, I found that it wasn't always a simple number of avatars present that could affect performance, but probably what they were wearing. One or two avatars entering a region could sometimes cause one or more seconds of delay to scripts running a nominal 1-second timer check. Once the region had assimilated them, the seconds lost stabilised. With fewer than 8 avatars present some of the lost seconds could be clawed back, but generally with 8 or more avatars in a region, lost seconds remained lost, and even slowly increased.

    You could therefore get some idea of how robust your system is by getting two or three people with loads of scripted attachments and getting them to enter and leave the region, sometimes one by one, sometimes all together.

    I'm not sure what your particular concerns are for your system, but it if's related to messages, I can put your mind at rest, over a 6-week period many thousands of test messages were sent with not a single missed or delayed incident, and some of the test regions were quite appallingly lagged. The avatar-entry spike didn't show up in the messaging test either. Sending and receiving messages is therefore not going to be an issue (although there is still thew classic delay on rezzing an object before it is available to talk to other objects)

    • Like 2
    • Thanks 1
  3. It will have to go in the listener script that is put into each object. Each object, when the script starts, will therefore have a channel derived from it's key, which is unique to it.

    The talker is going to have to do something else, it must somehow know the key of the object it is going to communicate with, and then it uses that same line of code to work out what the channel for the object must be.

     

  4. On 2/26/2022 at 3:03 AM, Coffee Pancake said:

    A more robust system that didn't leave the client hanging for a TP fail

    It's interesting you should say "robust", because for the past 9 months I've been exploring the "other" grids and have not once had a TP crash-to-desktop. Plenty of "destination cannot be found",  but never a fail with a crash to desktop. It could of course be the sparse population that gives a less onerous environment...

  5. Uh-oh, what have I gone and done?

     

    MY working assumption seems to be that of Nran,  my viewer will download all the objects, all the avatars, all their attachments, all the textures including bump and specular, but then the fun begins with how much I let it load to the scene on the screen. Have I told it to not bother with avatars above a certain complexity? Have I told it to not show bump and specular by turning off ALM, have I said I want a low draw distance , and do I want things to degenerate into triangles or not? My personal settings don't impact those around me.

    Back in 2010 there was a belief that somebody with a very high bandwidth setting would cause lag in those around them by grabbing all the server resource and hogging them. The misconception seems to have shifted away from the bandwidth slider to the LoD setting.

  6. 58 minutes ago, Jaylinbridges said:

    Quoting my first post Prof, I know where it is. 

    Oh gosh crikey, yes you did. I missed that. Maybe I need to go to the optician and ask for some glasses with a higher Lod factor different focal length :)

     

    59 minutes ago, Jaylinbridges said:

    Maybe I should correct the thread title to something more snarky

    There are days when I feel like changing my sig to a line from Screamin' Jay Hawkins, but let's not go there until we absolutely have to :)

    • Haha 1
  7. 13 hours ago, Wulfie Reanimator said:

    Why the snark?

    Since it's you asking, I'll give a sensible reply. (I am no boojum),

    It's just push-back.

    The existence of creations requiring a higher-than-normal LoD setting is well-known and irreversible, and while it is lamentable, it's not a game-killer. However, I recognise it does raise the hackles of several people who like to remind us, in case we forget, that it's the fault of creators who should know better. I, in turn, like to remind people that there are far worse sins.

    • Like 1
  8. It might be better to have a list of known URLs and select a random one from then instead of generating ones in the style of a thousand monkeys hoping to output a Shakespearean sonnet?

    The way I see it working is you would have to maintain a script with an initial list of various URLs, and on touching it, it would generate a random number between 0 and the number of URLs in the list less one, and present the chosen one to the toucher. Does this sound close to what you had in ind?

  9. To the OP - also try the debug setting RenderVolumeLod and give it whatever figure you choose. Jut don't tell anybody you're dong it.

     

    To a few others - I'm an old Laissez-Faire Liberal, I don't see why people can't put their Lod setting where THEY want it to be.

    Oh, I just realiesd, it would deprive the forums of a significant number of posts?

    Ok, Stet

     

     

  10. 9 hours ago, RilaVirum said:

    I want to be able to spin objects in any axis and then eventually by angle.

    To expand a little on Rolig's reply, the difference between using TargetOmega and  llSetRot/primitiveParams depends on whether you need everybody viewing the object to see exactly the same thing:

    A decorative windmill - use targetOmega because it doesn't matter where the different people see the blades

    A wheel of fortune pointer - use actual rotation because everybody needs to see the pointer in exactly the same place.

    • Like 2
  11. As Fluffy has said (whilst I was beavering away at the keys), you can't disable the right-click menu, but you can remove the touch part that an object may have inherited, which causes the cursor the change when hovering over the object.

     

    Create a new script inside it, it should then say "Hello Avatar"

    Edit the script and just after the line llSay(0, "Hello Avatar");

    add the new line 

    llSetClickAction(CLICK_ACTION_NONE);

     

    If there is a

    touch_start(integer num_detected) line followed by

    {

    then some code and a final

    delete that block.

     

    Save the script , make sure there are no syntax erors.

    Check that touching it does nothing.

    Delete the script.

     

  12. Dora's script is a little terse even when there are comments but it taught me quite a few things. Your thoughts about 

    23 minutes ago, Quistess Alpha said:

    Really the main insight I got is that for a randomly-generated path, you can probably make things 'prettier' by having fewer curve segments and more interpolated points per segment,

    parallel mine, I was impressed with how natural the paths seemed to be with none of the drunkard's walk effect that comes with just throwing random numbers in. Nature is chaotic, but very rarely random. (I believe the only truly random events are the decay of radio-isotopes, they cannot be predicted as individual events, only as statistical probabilities, hence their use for the infamous cat in a box thought experiment).

    • Like 1
  13. My RC parcel restarted first at 2022-01-26 at 9:09PM, before I logged in, and then while I was logged in but at another place I got a restart notice that it had done it again, at 2022-01-26 7:56 AM

    Somebody mentioned the Goons in the general forum, I'm now humming "I'm walking backwards to Christmas"

×
×
  • Create New...