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

[Mircea Lobo]

Since I finally managed to get Second Life running on Linux, I looked a bit into the recent things that happened in the SL community. I've been informed a while ago that Linden dropped OpenSim support due to a clause with an upcoming pathfinding algorithm. The licensing reason they invoked was so silly I'm not even going to comment on it... but I was curious to find out exactly what this is all about, and what was that important thing they cut OpenSim off for. So today I poked a bit more into this.

Someone clarified and explained that it's a feature which would allow primitives to navigate a simulator. And the client part (for which they "needed" to cut off OpenSim) is a library to preview the navigation paths on a mesh. I assume this means that when a person uploads a mesh to Second Life, he / she might be interested how a primitive using such an algorithm would know to navigate on it, and needs to see a preview of that.

Although I always like seeing new features put in, it's impossible for me to understand why Second Life would ever need something like this. Not to be rude, but I've been trying for an hour to think of a more useless thing SL could have than path finding support, and still haven't succeeded. Not only that I don't see why anyone would care for a primitive to navigate even between simple shapes (like blocks or cylinders), but meshes? I wasn't even sure if meshes have proper collision detection yet... since many games assign a bounding box and never bother to calculate these. There are modern games out there (like shooters) with cutting-edge graphics and technology, and I think even some of them don't have this.

Why do I consider this the most useless feature in SL? First of all, there isn't (and probably never will be) support for actual characters and bots. OpenSim tries that by simulating automated clients / avatars, and even so they have no use but walking around. IIRC this has never been a purpose in Second Life either, though things can change. While with a good LSL script you can make objects seem sentient, I don't see a circumstance where this would be done at large scale and needs such accuracy and detail.

Second, primitives can't even be animated yet! Sure, you can script child objects to change position and rotation (to simulate limb swinging for example), but that's not real animation and is very bad in all cases I've seen. They will always be simple and also very laggy, since the server needs to update them. The only way this could even remotely have an use is when Second Life will have self animated meshes, and LSL can set a model to loop an animation client-side (while walking for example). Priorities for LL are completely upside down IMO.

Third, there are so many things of huge importance compared to path finding, that even considering this feature is silly. Second Life still doesn't have local movement (you wait at least 0.5 seconds for the avatar to start stepping after you press the forward button). It has no collisions for water yet... no support for transparent + reflective (shiny) surfaces, and so many other things. I'd see more sense in taking a copy of Quake 1 and adding path finding support to that. There's such a big list of features SL truly needs for years that people would laugh to even hear of mesh navigation.

Those are just my 2 cents. I'm not looking to start drama, but I am pissed to see what LL decided to throw us OpenSim users away for. But other than that, my question is if anyone's ever going to use this feature at all. I've been a builder for years, but I think this is the first feature I'll never touch and can't find any real use for. If someone else can though, I'd like to see some concrete examples of what OS was sacrificed for.

[Czari Zenovka]

I haven't heard anyone I know come up with a use for pathfinding.  As I understood it (and I could be way off) it would allow for the same type of movement one finds with NPCs in MMORPGs.  For SL I could see this only having appeal for someone who wanted to create a game sim with roaming NPC objects.

The VKC dogs in Second Life are the closest to artifical intelligence I have seen here.  When pathfinding was announced I wondered how it would differ to whatever causes those dogs to roam seemingly independently as well as respond to text commands. (Not talking about the menu-driven type pets.)

IIRC, I read something on the forums about pathfinding essentially making many vehicles created prior to it not work any longer.  Anyone I know who has the ability to make estate changes has pathfinding turned off.

My personal opinion is that LL is releasing certain "features" in SL as a testing ground for the new games they are developing under the helm of our current CEO.

[Jadeclaw Denfu]

That's one purpose of pathfinding: http://maps.secondlife.com/secondlife/Jelloab/125/82/52

Collision avoidance. Watch the cubes avoiding the pillars, each other and you.
Yes, try to walk into their way, the cubes will go around you or turn away.

(If the cubes don't move when you visit, click them once to get them moving.)

 

[Czari Zenovka]

---

Innula Zenovka wrote:

---

Czari Zenovka wrote:

My personal opinion is that LL is releasing certain "features" in SL as a testing ground for the new games they are developing under the helm of our current CEO.

---

/me tries unsucessfully  to imagine how pathfinding could possibly be related to any of the new games that have actually been released or even announced, 

 

---

Rod basically said in an interview for a gaming website that he had been planning on starting his own company after leaving EA to develop games around "creative spaces" that "emphasize creativity tools more."  When he was offered the position of CEO of LL, after "laughing" and asking "is that still around" he "...saw a company that was ready made to do a whole bunch of other products, which I wanted to do."

Nothing would surprise me at this point.

[Innula Zenovka]

Yeah, but what I'm saying is that pathfinding doesn't actually have any connection with any of the games that LL have, in fact, developed.

I'm not completely sure, in fact, there would be any particular point in using SL as "a testing ground" in for a feature in a new game -- wouldn't it make far more sense to develop the platform for the new game from scratch (that is, what they actually  done with the new games they're releasing at the moment)?

[Czari Zenovka]

---

Innula Zenovka wrote:

Yeah, but what I'm saying is that pathfinding doesn't actually have any connection with any of the games that LL have, in fact, developed.

I'm not completely sure, in fact, there would be any particular point in using SL as "a testing ground" in for a feature in a new game -- wouldn't it make far more sense to develop the platform for the new game from scratch (that is, what they actually  done with the new games they're releasing at the moment)?

---

One would think.  And in the interview Rod did say he had planned on starting his own company but he found a "...company that was ready made" to do some of the things he wanted to do.  I could be out in left field and I did preface my comment with "My personal opinion is..."

[Orca Flotta]

One good use for pathfinding would be in Ann Otoole's robot cars that cause so much grief on mainland roads. But since nobody likes or rides them anyway why bother? I couldn't think of any other reason.

Yeah, Rod Humble abuses his position, his employing company, the labour of his underlings and LL's funds (they receive from us) to develop his private obsession. It's like we SL residents are financing the kickstarter for his own company. He can't be bothered with running the daily business of SL. Wrong guy at the helm of LL.

[Innula Zenovka]

It's AnnMarie Otoole who makes the cars, not Ann.   I don't know AnnMarie's about the matter, but I think the frequent and understandable confusion between the two is a bit of a sore point with the latter.

I'm a bit surprised you say Rod Humble can't be bothered with running the daily business of SL.  They recently started hiring again, and are advertising several senior SL-specific posts as well as ones aimed at new products.

[161488303349]

---

Jadeclaw Denfu wrote:

That's one purpose of pathfinding:

http://maps.secondlife.com/secondlife/Jelloab/125/82/52

Collision avoidance. Watch the cubes avoiding the pillars, each other and you.

Yes, try to walk into their way, the cubes will go around you or turn away.

(If the cubes don't move when you visit, click them once to get them moving.)

 

---

^^ that ^^

one thing that brings sims to their knees the most is collisions. the next most is running detectors to avoid collisons

if we can have an easy way to not have to do either then is better in the long run

yes we want all the other things that come wiht NPCs. but if the base navigation/avoidance isnt in place first them will just make the existing/old script/physics worse if is no navmesh system

is early days as well. but i think linden are doing this in the right order

+

also the Havok licensing is the problem for OpenSim. not linden

to please OpenSim then linden could dump Havok altogether and go with another engine. have to rewrite the whole SL physics system to make that happen tho

 

 

[Amethyst Jetaime]

Right now I don't know of anyone using and like some have posted everyone with a sim I know has turned it off.  I wish LL had made it off by default rather than on.

I could see a use for automated vehicles, wildlife, or for background  "people" in a role play sim or maybe bartenders in a club to contribute to atmosphere, however animating these things still is a problem as they still will look like robots.  The other issue is that you need to make these things as well as  paths for them to follow, which uses up prims, so again it won't be practical for a lot of sims.

Personally I think that LL hoped people would devise more video type games like shoot em ups so they'd attract more new people and LL could make more money.  But I don't see an economic model that would work for a resident to do that unless they charged admission, and we all know that people in SL won't pay admission and aren't into gaming of that type in SL. They log out of SL and into an on line gamed, which can do those things much better.

For right now, I see pathfinding is a new feature that no one wanted or asked for and that has used resources and time that could have been used for developing features that we actually could use and want, such as a better mesh introduction and deformer, to say nothing of fixing things that have been broken forever. 

[Orca Flotta]

And there I was thinking foolishly that AnnMarie was an alt of Ann, particularly that I see now some robot  cars made by another AnnMarie with a different last name. Most of those cars look like they were originally made by Arcadia Asylum anyway.

And I say that Rod Humble can't be bothered because I see the direction in which he drives the company. Yes yes, I admit they fixed the rendering and made V3 almost as nice as a TPV but still I guess he's on his own agenda. Didn't know they started hiring again, sorry.

[Jadeclaw Denfu]

Orca Flotta wrote: "particularly that I see now some robot  cars made by another AnnMarie with a different last name."

Yep, AnnMarie Oleander. That is an alt for AnnMarie Otoole. When finding one, report them to AnnMarie Otoole.
Remember, always report them, when you find one, where it doesn't belong.
Often, you'll find multiple cars stacked onto each other. Or parts strewn around. Or both. Or a whole junkyard.

The problem with these things is, it seems she doesn't want to fix the programming, as a lot of the problems can be fixed.
I've had some discussions about that with her, suggestions I made are using waypoints to keep these cars on the road
and use them to set exclusion zones and better filtering of position data, which would stop the cars from shooting into
banlines. Yes, at one crossing, while I was writing a message about one car, I watched a bus turning into one sideroad,
going slowly up that road, only to see that bus coming back 3 seconds later at 500mph, ending up on a parcel behind me,
stuck inside banlines.

Pathfinding should make it easier to avoid stuff like that and make the behavior of automated objects more predictable.

 

 

[JobsFox]

Pathfinding is FAR from useless. In fact it solves some HUGE problems for product developers and estate managers in SL.

Snuffles is the first breedable that uses the pathfinding system. They generate hardly any lag because they do not use old collision based movement systems. 

http://www.goldtokens.net/index.php?title=Snuffles

Just because some people so far haven't had the technical knowledge of imagination to make something useful out of it does not make it useless.

We run the probably the largest gaming network in Second Life. We develop new games used by 30,000 players per month. I can say from experience that game development is something that takes MONTHS. So there is probably a ton of wonderful pathfinding content still in development that people just haven't seen released yet.

Actually the real problem is that we need to make pathfinding more accessible to users. It should be possible to make a prim into a walkable surface or exculsion volume just using a dropdown box on the EDIT menu instead of having a scary looking linkset menu. Also, when someone makes a new walkable surface or exclusion volume it should rebake the navmesh immediately. Most users understandably lack the technical knowledge to perform such tasks.

[Mircea Lobo]

Well, the part about automated vehicles would make some sense. Although in all the sims I've explored during the 7 years I've been with Second Life, I can't remember seeing any attempt to make automated vehicles anywhere (except maybe some trains). In the sense that it didn't seem like a feature anyone thought about, not because they couldn't do it.

For animals or other creatures it make no sense. Since like I explained, there is no way to animate them. I seen some attempts to use LSL for animating limbs on some prim creatures, but that is both very limited and very laggy (since it's basically the server animating and sending animation changes). The only way objects could ever be used for creatues is when self-animated meshes will also be supported, which IMO should have been LL's priority instead of this.

For either of the two cases however, making objects navigate a sim should have already been easy and possible. One way is: The sim owner could place invisible primitives as nodes on the ground, then a script inside the object would know to look for the nearest one in the chain and navigate there. Second way is not even using nodes, but tracing toward a location to see if there's anything there in a LSL script. But this would be more difficult and I don't know how possible it is exactly. Surely this was doable before LL wasted their time to add path finding however.

[Innula Zenovka]

To my mind, pathfinding hasn't been around long enough to make much of an impact.   It's only a few months since it's been available on the main grid at all, and there was a fair bit of bug fixing to do after that.   In my experience, certainly, people take time to explore the possibilities of new technologies like pathfinding.    SL is a recreation for most of us, certainly, and there's only a small minoritiy who are ever going to have both the ability and inclination to experiment with making pathfinding content, and certainly initially.   Give it a year or so, and things might look a bit different.

LL, I think, scored an own goal with the way pathfinding was introduced.   It certainly didn't help that it was rolled out at the same time a major (but unconnected, and now fixed) sim bug was introduced, so a lot of people turned off pathfinding on their sims "in case that makes a difference" (it didn't).    

Furthermore, initially, at least, LL's documentation about what sim owners needed to do to optimise their lands for pathfinding made the whole process sound far more complex than it need be (or so I'm told by people who have optimised their lands) and certainly overstated the performance hit you were likely to take if you left it on and didn't optimise for pathfinding.  

I turned it off on all lands for which I'm responsible, which I now think, based on what I've read since and heard from people who've experimented with having it on, was probably unnecessary,  but it's still turned off on my lands because I haven't got round to doing anything about optimising them (still not a trivial undertaking, though not as bad as initially I feared).  

Meanwhile, although my tenants know that I will make the time to optimise a sim if asked, they have a disincentive to experimenting with pathfinding products because they know they won't be able to use them at home immediately and it means inconveniencing me before they can use them.   In turn, that means I've no particular reason to make time to optimise a sim when there's plenty of other things I either need or want to do with my time in SL.

 

[Qie Niangao]

Yes, I think wandering breedables and game-enhancing critters were the intended first applications of Pathfinding, and it's not surprising there are increasing numbers implemented that way now.

Nonetheless, I'm kind of disappointed in Pathfinding. Admittedly, it's a failure of imagination on my part, but I kept casting about for some non-critter, non-NPC application of it, and coming up with nothing. It's really not all that useful for large-scale automated vehicles (such as AnnMarie's); among other reasons that's because, in the uncharted wild, the navmesh is full of traps too intricate for Pathfinding to navigate. 

On the other hand, some related technical features--notably keyframed motion and raycasting--are immensely useful for a very wide range of fun new stuff that will add a lot to Second Life. Probably that's to be expected: the more fundamental a technology, the wider its potential applicability--and the more effort required to develop on it.

Oh, about possible tie-ins with new Linden products: I've been thinking that NPCs (with or without Pathfinding) might use whatever they're now calling the storytelling AI technology they acquired with Little Text People. No idea if there's any LSL-relevant API in the works, though.

[Innula Zenovka]

---

Mircea Lobo wrote:

I seen some attempts to use LSL for animating limbs on some prim creatures, but that is both very limited 

---

Limited to things like Psion Chickens, Amaretto Horses, Ozimal Bunnies, Meeroos, all manner of dogs and cats.....

Laggy or not, these things are very popular -- in SL at least, though maybe not in Open Sim --  and pathfinding offers a far less laggy (and far simpler)  way of making them.

[Mircea Lobo]

I've seen animations done that way years ago actually... like pets that roam around sims. But in my opinion that's a bad way of animating anything, and only used because it's currently the only way. IIRC the only animation function that works client-side at all is llTargetOmega. But you give that a constant direction, so you still need the server to update it every few milliseconds so limbs swing back and forth properly. This will always be laggy, no matter how good one's connection to the server is (not to mention you can't bend limbs and do many other things you can with armatures).

I just find it bad priority ordering to add a feature that SL could perfectly live without, without adding mesh animation support first to even give it a complete purpose. It's like adding support for guns in a game where you don't even have players who can connect yet. That's just my opinion of course... I've been with SL for years and also in the development of many games, so I can only hope my vision on these things is usually correct.

[Innula Zenovka]

But what I'm saying is that, despite the fact that existing ways of getting pets to roam round sims are both laggy and complicated to implement, people do, nevertheless, make and buy these laggy pets in large numbers (often to the fury of their neighbours).    

That being the case, doesn't it make sense for LL to provide an easier and less laggy way for people to do what they're going to do anyway?  

 

[Mircea Lobo]

I didn't know people were really interested in object pets and roaming primitives. Anyway, I imagined the LSL way wouldn't have been too laggy, though I never tried to do path finding in a LSL script. Other builders could have made a free navigation script and distributed it for free too, so other artists wouldn't worry about creating one themselves. Also, if some players wanted creature objects, I once again believe that the first thing LL should have given them is support for self-animated meshes (then fine, the path finding thing).

I'm also surprised this is one of the first things LL decided to give attention to, in terms of major features. There are people who are suggesting important changes to them for many years, and for some the Lindens barely bothered to reply. For example, I've been wanting support for transparent + shiny surfaces for 5 years, which should be a cinch to implement at this day... yet such a basic thing still doesn't exist. Knowing LL they also never cared what the community wanted, and just went for ideas *they* considered are need (eg: Everyone hated web profiles but they still shoved them onto us and won't change their minds).

[Innula Zenovka]

I don't know if this is going to provide the transparent + shiny surfaces you want:  http://community.secondlife.com/t5/Tools-and-Technology/New-Open-Source-Project-to-Improve-Graphics-Rendering/ba-p/1637559

 

[Paul Hexem]

What I get from this forum thread is "I can't think of a way to use Pathfinding and I don't like products that do, so it's a useless feature! MY feature requests are FAR more important!"

[Masami Kuramoto]

---

Mircea Lobo wrote:

I am pissed to see what LL decided to throw us OpenSim users away for.

---

LL's move should be seen as an opportunity for OpenSim. For example, due to compatibility no longer being a concern, OpenSim could roll out the mesh deformer ahead of LL.

[161488303349]

now that Firestorm have decided to fork their viewer into SL and OS versions. i think that OS (or a fork of it) is finally going to make the break and cut the linden apron strings

for the future of VWs overall i think this is a good thing

[Czari Zenovka]

---

Amethyst Jetaime wrote:

 

Personally I think that LL hoped people would devise more video type games like shoot em ups so they'd attract more new people and LL could make more money.  But I don't see an economic model that would work for a resident to do that unless they charged admission, and we all know that people in SL won't pay admission and aren't into gaming of that type in SL. They log out of SL and into an on line gamed, which can do those things much better.

For right now, I see pathfinding is a new feature that no one wanted or asked for and that has used resources and time that could have been used for developing features that we actually could use and want, such as a better mesh introduction and deformer, to say nothing of fixing things that have been broken forever. 

---

Agree on both counts.  It isn't much of a stretch to imagine LL instituting pathfinding in hopes of attracting the gamer community, especially being introduced at approximately the same time as LL's being listed on Steam.  It is fairly obvious (gleaned from blogs of several well-respected SL bloggers), forum chat, and personal observation/opinion that Rod would prefer to change SL into a game platform.

Nothing to add to the second paragraph.  Well said.