Jump to content

Why I consider "path finding" useless in SL, and what do people use it for?


Mircea Lobo
You are about to reply to a thread that has been inactive for 2864 days.

Please take a moment to consider if this thread is worth bumping.

Recommended Posts


Innula Zenovka wrote:


 What sort of errors are these that you've seen in point 3 that require a region restart?   All I've seen are messages that the Navmesh has pending changes, which anyone with building rights can force the region to commit (or which the region itself commits if you leave it for 10 minutes or so).   See
of the wiki article on Pathfinding Tools in the Second Life Viewer.   What icon are you seeing that you interpret as indicating "pathfinding errors"?

Everything I've read, you see, leads me to think the ideal state of affairs would be have pathfinding turned on, having first optimised the region for pathfinding.    The disadvantage of having pathfinding turned on in a region that's not been optimised is that the pathfinding objects won't work as well as they should but won't, or shouldn't according to Maestro Linden (who normally knows what he's talking about) , cause any noticable hits to sim performance.   

That's why I'm querying your statement that "most sims I see would probably be best with pathfinding disabled".   I would agree with "most sims I see would probably be best with pathfinding optimised" but I really don't there's a huge disadvantage to having it turned on.

Someday I will agree with you that optimizing for pathfinding is better than turning it off. Just not yet. Right now, when you're terraforming, you are only moving the land, not the navmesh. When they don't match, that's when you get the error. On firestorm, if you look on top of your screen, the S icon is pathfinding. (really it's a wavy arrow.) That causes an annoying flash, and since pathfinding is basically a "second ground" it's un-needed information if you do not have moving objects on your land.

Even still, for pathfinding to serve any purpose whatsoever, it has to be on land that will be used for such a purpose. Many of the private regions do not allow scripted animals at this time. (Many do, so those ones are not the ones I'm referring to.) Maybe in the future, as pathfinding improves, it'll cause those that do not allow scripted animals to change their minds, but until those sims allow those things that'll use pathfinding, it won't help them at all.

*edited to add* Does anyone else get the impression that if everybody on here liked Opensim as much as they say they do, it would be easier to find people there?

Link to post
Share on other sites
  • Replies 287
  • Created
  • Last Reply

Mircea Lobo wrote:

The only OpenSim grid I'm actually familiar with is OSGrid. Not sure if it's the largest, but it's surely ran by the core developers too. I don't know how often they log in to build stuff, but I know some have their own sims and did create new content (Nebadon for instance has many good creations there). But I guess it depends on what the owner of each grid cares to do... some might only be interested un running the grid, others might be users who go there themselves.

I think we're talking at cross purposes here. I'm not talking about creating objects. I'm talking about creating the system. The creators of the OS system are able to continue creating the system, whether it follows SL or not. The owners of grids that use the OS system are unable to create something different to SL.

It came about when we (the thread) were talking about OS going off on its own and not merely mirroring SL. I said that the OS grid owners (not the OS creators) are unable to do that. They are stuck with someone else's system and all they can do is (maybe) tweak some parameters.

Link to post
Share on other sites

Phil Deakins wrote:

 

>> "The point is that OS grid owners cannot create. They merely use the creation of other people - the OS creators."

 

There is no such separation between OS grid owners and OS creators.

 

OSgrid: run by OpenSim developers. Inworldz: maintaining a proprietary PhysX-enabled fork of OpenSim. Avination: founded by Melanie Thielker, OpenSim developer who contributed to the Hypergrid protocol. ScienceSim: run by Intel Labs, developer of OpenSim's Distributed Scene Graph. Kitely: created proprietary OpenSim extensions to improve performance for their cloud-based region hosting.

 

Many OpenSim developers run their own grids, and any grid owner can make modifications to OpenSim's core or write extension modules. The fact that some grids are run by non-developers doesn't make your statement right. Every grid owner _can_ create if they so choose, and many of them do. That is the whole point of having an open source server.

 

Linden Lab just made it very clear that they don't want to play a role in any of this any more. Their Architecture Working Group is dead. Their own technology is becoming irrelevant. The Phoenix team just took the crown in four categories: most popular SL viewer, most advanced SL viewer, most stable SL viewer, and most OpenSim-friendly SL viewer. LL is trying to restrain them by tightening the Third-Party Viewer Policy, but it doesn't work. LL keeps adding stuff that no one wants or needs (web profiles, pathfinding) or stuff that just plain doesn't work (direct delivery) while blocking the stuff that everyone is waiting for (e.g. the mesh deformer).

 

LL's only remaining asset is the content that people have uploaded to their grid over the past eight years. This is what keeps their users locked in and paying tier through the nose, while the Lab packs more and more regions on less and less physical hardware. They will milk this cow until it runs dry, and then throw away all your "virtual property" in a heartbeat.

Link to post
Share on other sites

have you seen the newest Cloud Party?

if SL ever close its doors, thats where we all going

why?

bc we not have any stuff like you say. and we not going to spend another fortune to get the same stuff back to do the same stuff all over again

and Cloud Party run on a browser and on a tablet pretty good

and Unity worlds as well

like CP does Unity got way better tech than OS/LL will ever have

 

 

Link to post
Share on other sites


Masami Kuramoto wrote:

. The Phoenix team just took the crown in four categories: most popular SL viewer, most advanced SL viewer, most stable SL viewer, and most OpenSim-friendly SL viewer. LL is trying to restrain them by tightening the Third-Party Viewer Policy, but it doesn't work. LL keeps adding stuff that no one wants or needs (web profiles, pathfinding) or stuff that just plain doesn't work (direct delivery) while blocking the stuff that everyone is waiting for (e.g. the mesh deformer).


why do you brag the TPVs up like this just to try make some obscure point that opensource is somehow better than closed

the Phoenix team devs are serious people, same all other the TPV devs. they know that 80%+ of the viewer code is written by linden devs. the TPV dev got respect for them. they dont need people like you dissing the linden devs in your rants about this

+

also about the deformer

right at the very start a lady asked: what about the buttons?

she still waiting for the answer after nearly 2 years now

the serious devs on all sides who actual know what this question means dont know what is the answer yet. until they do the mesh deformer is practically useless and will never be ready for consumer grade release until the answer is found. if it ever can be

the non-serious devs and their camp followers dont even know what the question means 

 

 

Link to post
Share on other sites

Oh, that yes. Thankfully the SL viewer is open-source, and continuing on the path of a custom OpenSim viewer will likely be the way. Some people are still thinking about an original viewer written form scratch, but IMO there's no way that anyone will make one in a decade, as good as SL and with the same features.

In this sense, maybe it is better OpenSim separated from SL... we might have the possibility to become even better now, and a more different alternative than the original SL. Even though I never wanted OpenSim to become a competition for SL... I always just seen it as the "fully free Second Life".

Link to post
Share on other sites

You didn't pay attention. Linden Lab hired a contractor to develop that viewer. Many of the Lindens who developed the original viewer were laid off long ago.

 

Addressing the question how the deformer will handle buttons: not at all! Objects of this size are supposed to be part of the texture. Modelling them is a stupid idea.

Link to post
Share on other sites


Masami Kuramoto wrote:

Phil Deakins wrote:

 

 

Linden Lab just made it very clear that they don't want to play a role in any of this any more. Their Architecture Working Group is dead. Their own technology is becoming irrelevant. The Phoenix team just took the crown in four categories: most popular SL viewer, most advanced SL viewer, most stable SL viewer, and most OpenSim-friendly SL viewer. LL is trying to restrain them by tightening the Third-Party Viewer Policy, but it doesn't work. LL keeps adding stuff that no one wants or needs (web profiles, pathfinding) or stuff that just plain doesn't work (direct delivery) while blocking the stuff that everyone is waiting for (e.g. the mesh deformer).

If you look at the latest Firestorm release notes, I think you will see that the changes listed as "Linden Lab New Features and Improvements" are by no means insignificant.   You could well argue that TPV devs take LL's work and improve it (primarily by changes to the UI) but I think it's undeniable -- certainly it's what every TPV dev I've ever spoken to says -- that well over 95% of the viewer is, and always will be, LL's work.   

I'm a bit surprised, by the way, to see you describe either Phoenix or Firestorms as the "most OpenSim-friendly SL viewer".   I'm not disputing it, but in what ways would you say they're more "OpenSim-friendly" than Kokua/Imprudence, which have always been made with OpenSim very much in mind, as I understand it?   I also get the impression Henri is very aware of OpenSim when he makes Cool VL.

Link to post
Share on other sites

We seem to have gone off on a Tangent from the main topic here again, Second Life Versus Other Grids.

So as long as we are on that tangent, I want to add a thought.

If SL was to shut down, I think many folks would move to another Grid.  And then the folks running those Grids would discover it's not as inexspensive to do as it has been for them so far.

It's a simple phenomena that we see in RL all time.  As population density inceases, the cost per individual to maintain infrasturcture goes up, not down.  That is why taxes in the city are higher.  After all, how often have we heard the phrase, "It's cheaper in the country."

Just food for thought here.

 

 P.S. Sometimes I wish Lithium would let us post replies without it looking like we were responding to some one.

That is, this is just a general comment.

Link to post
Share on other sites


Masami Kuramoto wrote:

Cloud Party's business model isn't different from Linden Lab's in any way. They let you produce free content for their platform and then make you pay for the privilege of using it. You will be locked in again.

yes

the thing is that SL is not about the stuff. not for most of us. is about the inworld and the people in it. we are SL. is deep ingrained in our pixels. in our RL minds and hearts as well. is our home. many of us think of our pixel/digital/avatar selves as Sansaran

if our world implode then we will scatter to places where we not reminded of our home world. we will go to new worlds and make new lives and new ways. we will remember the past and treasure it and go on to be bold and strong again in a different way

unless a OS based grid can help us do this, to be differemt, then we will not come just to recreate/replicate what we have lost. is irreplaceable that part is

 

Link to post
Share on other sites

In RL, some of us -- many of us -- prefer the convenience and economic advantage of living in the city rather than the countryside, and are prepared to pay the extra costs associated with that.       To my mind, virtual worlds are much the same.   

I am always, I have to say, somewhat sceptical about claims for the technological abilities of the non-LL grids.   It seems to me that many things are a lot simpler if you've only got, at most, a couple of hundred avatars online at any one time, rather than several tens of thousands 24/7.

Link to post
Share on other sites


Masami Kuramoto wrote:

You didn't pay attention. Linden Lab hired a contractor to develop that viewer. Many of the Lindens who developed the original viewer were laid off long ago.

 

Addressing the question how the deformer will handle buttons: not at all! Objects of this size are supposed to be part of the texture. Modelling them is a stupid idea.


the hired company wrote a V2 UI that overlay the linden core. same like the TPVs do now. is cosmetic work. most of the work that goes into TPVs is the same. it surfaces functionality that already in the core. like thru UI buttons, menus and widgets and dialogs

and in doing this the TPVs also pick up on bugs and mod/fix them

in terms of substantial additions to core functionality can add them up:

RLV. Area Search. Spell Check. Step Render. Multi Attachments

these all good things. but in the great scheme of the entire codebase is not all that many

+

we already got buttons painted on mesh. has been that way since the beginning

the whole point of mesh clothes  is to not do painted on clothes

if the button problem is not solved then we going to end up in deformable sack cloth. and if that is all it ever will be then linden can make a fairly straightforward mod to the existing system clothng/hair dialogs and we be able to make sack cloth that deform perfectly

the way that is done is by having a cutting scissors tool that lets us cut the avatar mesh itself and pop it out a bit then paint/texture on the cutout. and use a smoothing iron to blend the edges

 add in an extruding tool for hair as well and we done

edit: and a curling iron please

+

btw the button problem is actual unsolvable for the general case. if anyone ever did solve it then they will simultaneously solve at least 2 of the Millenium Prizes for a $1,000,000 each. and break every known computer encryption algo

 

Link to post
Share on other sites

ps

i just add to here

about a mod to the clothes dialog. if i ever made a grid then i would do this way. so can have a pretty easy way for users to make their own mesh clothes inworld

then i make so that another mesh object made independently like buttons, buckles, shoelaces, zippers, etc can be linked/attached to a vertex on the sack/body mesh. so when the avatar sliders are moved then the buttons etc move in and away and not deform automagically. then just like prims/sculpts do now they can be resized by themselves

Link to post
Share on other sites


Masami Kuramoto wrote:

 

Addressing the question how the deformer will handle buttons: not at all! Objects of this size are supposed to be part of the texture. Modelling them is a stupid idea.

You obviously don't own the same line of mesh shorts as I do.

I think I have them from a couple of places.

- One of them has modeled buttons, and the difference is striking.

 

Not to mention such things as drawstrings, lace and frills, things coming out of pockets, and so many small details that when modeled, are just amazing to see.

 

 

Link to post
Share on other sites


16 wrote:

 

also about the deformer

right at the very start a lady asked: what about the buttons?

she still waiting for the answer after nearly 2 years now

the serious devs on all sides who actual know what this question means dont know what is the answer yet. until they do the mesh deformer is practically useless and will never be ready for consumer grade release until the answer is found. if it ever can be

the non-serious devs and their camp followers dont even know what the question means 

Oh the sourceof that other comment.

Having worked in Poser back in the day with Wardrobe Wizard - the question is obvious to me. Did I ask it or did someone else, cause I think I asked something -like it-.

And yes to your answer...

This was a constant nightmare back in my Poser days when using deformers to fit an object made for one figure to another... Sometimes there was no solution and whole projects would have to be done over with different elements. Othertimes, you'd need heavy postwork, or to just export the whole thing as a 3D model into some unrigged format for Carrara / 3DSMax / etc and remodel portions...

- That last solution being too much to ask for hobby art.

 

I envision a future where people will want 'magnet zone deforming' like Poser, to correct what the deformer has deformed... but this only working to a limited degree... and the processing and bandwidth costs of it all will be insane for something streamed online.

That's why I just looked at the standard sizes, chose the one closest to my then shape, and made myself fit it - finding my uniqueness in my avatar's face. Even -WITH- a deformer, clothes not using it might come to be preferred by people wanting higher than 1fps... :D Not to mention the buttons... and the horror of what people will say when they come to understand what that question means...

And of course, they will -ALL- blame LLs for it, when its just a problem of 'nature of the beast.'

 

Link to post
Share on other sites


Masami Kuramoto wrote:

The hypergrid _is_ the inexpensive countryside. An unlimited number of small grids, with avatars traveling between them, thanks to a common network protocol. Decentralization is the key.

My point was would it and could it remain so inexpensive should it see a huge influx of avatars.  I don't think so but that is only a hypothesis based on practical observations.   The question really would be how much the expense would increase.  To somebody somewhere, in particular to the individual SIM owner.

And how about to a grid like InWorldz?

Link to post
Share on other sites


Pussycat Catnap wrote:

 

Oh the sourceof that other comment.

Having worked in Poser back in the day with Wardrobe Wizard - the question is obvious to me. Did I ask it or did someone else, cause I think I asked something -like it-.

 

 

it might have been you. i just dont remember exactly now who said it. at the time i read lots of forums and blog articles about it all. you was in them debates as well i remember

as soon as i read what about the buttons? then i went: yes thats right. like bing a light go off in my head

is only something someone who into fashion would know to ask instinctively even if they dont know nothing about the techy side

so i end up read up on mesh and 3D graphics and tools forums and then end up read heaps oif computer science papers as well and try to understand the math and the algos. final end up learn about kolmogorov complexity and P v NP and all that

+

just want to say that i think Qarl is a good person and talented codey. i also think he just underestimated the problem at the start. i think he knows this now tho

i only say underestimate bc right at the start he only ask for one binary flag to distinguish deform from non-deform mesh and that he will sort it out from there

after only a short time he then start to say: well actually can only do this much or that much or some other much. now the whole thing been parked up at linden. is some other people like Darien taking it forward now. just trying to do what they can. which is good. Nalates report on how they going at dif times. so thats good as well

+

now that i read up on it all and understand it better. i personal think that the whole approach that was taken is not the best path

i know what i would like and i think i know what might work and be acceptable to someone else like me. but i am not a maker so is not up to me to decide anything really

 

 

Link to post
Share on other sites


16 wrote:

we already got buttons painted on mesh. has been that way since the beginning

the whole point of mesh clothes  is to not do painted on clothes

Who said so? You?

The whole point of normal maps is to make painted buttons appear like 3D geometry. If your game engine does not support normal mapping, you bake ambient occlusion and/or specular reflections into your texture to fake a 3D appearance of the buttons. Only complete noobs try to import high poly clothes with mesh buttons. These people deserve to get screwed by the mesh deformer, because they don't understand how to optimize mesh assets for realtime rendering.

Link to post
Share on other sites


Masami Kuramoto wrote:


16 wrote:

we already got buttons painted on mesh. has been that way since the beginning

the whole point of mesh clothes  is to not do painted on clothes

Who said so? You?

The whole point of normal maps is to make painted buttons appear like 3D geometry. If your game engine does not support normal mapping, you bake ambient occlusion and/or specular reflections into your texture to fake a 3D appearance of the buttons. Only complete noobs try to import high poly clothes with mesh buttons. These people deserve to get screwed by the mesh deformer, because they don't understand how to optimize mesh assets for realtime rendering.

lol

is going to be really good when a unnoob awesome wonder codey like you can write the algo that can sychronise the deformation of two or more topologocial surfaces with disjoint vertices in real time

the world is waiting. you going to be famous when you done it. and rich. you willl even get a Nobel Prize. and if you under 35 then we give you a Fields medal as well

 

 

Link to post
Share on other sites

ps

about the normal map on a single topological surface. can you write the algo to deform that as well please. should be way easier than doing two or more simultaneously

also could you write the algo that stitch multiple surfaces with disjoint vertices into a single surface. that be good as well

after you done that can you stand with the light behind you

thanks

Link to post
Share on other sites


16 wrote:

lol

is going to be really good when a unnoob awesome wonder codey like you can write the algo that can sychronise the deformation of two or more topologocial surfaces with disjoint vertices in real time

the world is waiting. you going to be famous when you done it. and rich. you willl even get a Nobel Prize. and if you under 35 then we give you a Fields medal as well

On August 31st, 2011, I wrote this in a post to the SLU forum:

"I've seen the other proposal to shrink-wrap mesh attachments around the avatar, but my experience with similar functions in Blender tells me that it's not going to work. Mesh clothing is not just a simple layer that floats above the skin. It has wrinkles, folds, collars, appendices, rigid attachments. The idea that all these could be handled properly by a uniform deformer cage is unrealistic in my opinion. Deformer cages are meshes too, and they need to be tailor-made for the target mesh. Even worse, they need to be rigged as well. This is a very complex problem to solve, much more complex than mesh import itself, and the latter already took the Lab two years to complete."

Fast forward 15 months, the deformer still hasn't been rolled out. People consider it not ready for prime time because it has trouble with wrinkles, folds, collars, appendices, and rigid attachments.

The sound you are hearing now is the world's saddest song played on the world's smallest violin.

Link to post
Share on other sites


Masami Kuramoto wrote:

Fast forward 15 months, the deformer still hasn't been rolled out. People consider it not ready for prime time because it has trouble with wrinkles, folds, collars, appendices, and rigid attachments.

The sound you are hearing now is the world's saddest song played on the world's smallest violin.

The reason people don't consider it ready for prime time is because a certain person hyped a deformer up to be the best thing since sliced bread, made of magic pixie dust and unicorn farts. If that person hadn't claimed that a deformer would bring world peace then maybe people would have somewhat more realistic expectations of it now.

Link to post
Share on other sites


Masami Kuramoto wrote:

 

On August 31st, 2011, I wrote this in a post to the SLU forum:

"I've seen the other proposal to shrink-wrap mesh attachments around the avatar, but my experience with similar functions in Blender tells me that it's not going to work. Mesh clothing is not just a simple layer that floats above the skin. It has wrinkles, folds, collars, appendices, rigid attachments.
The idea that all these could be handled properly by a uniform deformer cage is unrealistic in my opinion.
Deformer cages are meshes too, and they need to be tailor-made for the target mesh. Even worse, they need to be rigged as well. This is a very complex problem to solve, much more complex than mesh import itself, and the latter already took the Lab two years to complete."

Fast forward 15 months, the deformer still hasn't been rolled out. People consider it not ready for prime time because it has trouble with wrinkles, folds, collars, appendices, and rigid attachments.

The sound you are hearing now is the world's saddest song played on the world's smallest violin.

good

we now both agree that the approach taken was not the best way to do it

is why is in the linden bin and i dont think is ever coming out unless is some massive breakthru in computer science happen

or a whole different approach is taken to make a kinda okish some other solution that people will go oh! ok that will do

+

i just also want to say that i am not a die in the ditch open v closed source person. each have its strength and weaknesses.

quite a lot of people in both camps depending on what is the specific thing they building/working on. or even what it is and what use it maybe has for them. lots of the biggest opensource contributers work on closed by day so they can get paid to feed themselves. and work on opensource in their evenings and free time. some are even lucky to be able to get paid to work on opensource by day

+

i not write a lot of actual code myself. but i do understand algos. well at least them that are written down in a paper. most times anyways

also whatever rare code i do/can write i actual release public domain. bc they mostly just puzzles for me and i like to know what is the answer. if there is one. i even done this in SL

for example somewhere way down in this forum is a arithmetic feistal network thingy i made in LSL, just for fun. i not invent the algos in it. i just code them

in it there is a very similar problem to what we discussing. like how to find the integer factor closest to sqrt for all integer N upto MAX_INT

my simple solution for all N upto MAX_INT was to brute force attack it.. and code my attack with the best known brute force attack algo that not use any lookup space. is faster algos but they use lookup table/space. the whole point of the script puzzle overall was not to use any lookup space to create "random" arrangements of permutations for all N upto MAX. like no lists or nothing

+

brute force attack is actual what Qarl did in his deformer code. he admit this on his first code release. he also said at the time that linden probably going to have kittens when they see it

since then the code has been tidy up a lot. other people done some tuning and modding as well. so is all good. but underneath it all, is still a brute force attack. and thats the way way way down deep problem

Link to post
Share on other sites

Archived

This topic is now archived and is closed to further replies.


×
×
  • Create New...