Innula Zenovka

(I'm posting this here because it affects everyone who might be buying mesh furniture or vehicles, not just content creators, but please move it if it's in the wrong place).

Via Nalates: #SecondLife Land Impact Costs to Change

In summary, it seems that the old trick of making a child prim phantom by setting llVolumeDetect(TRUE) and then linking it is going to stop working very shortly, when Pathfinding is rolled out. See PATHBUG-69.

However, this is to be done in a way that won't break much existing content (or shouldn't) because stuff's going to convert itself automagically:
Hand-in-hand with this comes a new LI accounting system, which Falcon Linden explained at the Simulator User Group yesterday:

__________________

Darien Caldwell

Sounds like a positive change overall. the way scripts affected LI is currently a bit crazy. Glad they toned that down a bit.

__________________
[H]arsh Styles

Innula Zenovka

As I understand it, the new system means, or will mean, that an object is never going to be more expensive, in terms of LI, than it would be if you calculated it by prim count, and will usually be cheaper. And how much cheaper is determined, in part, by how few scripts it contains rather simply by whether it's unscripted or scripted.

This sounds a very positive development. It'll certainly remove most of my concerns about using mesh in builds.

Void

don't like the land collision changes to volume detect.... makes object scanning more of a PITA. if an object is buried in a crevice (effectively underground) you only need to get close to sense it, which could be done underground. now you'll have to track land contours precisely which is expensive, slow and unnecessary for most general cases.

I'm sure there's other things it breaks, I remember someone made a tennuler vehicle that let avs travel underground, I'm betting it'll break too.

__________________
- These eyes can do more than see

Pandorah Ashdene

Does anyone know/understand if that cap only holds for legacy prims (cube, cylinder, tori, etc.) or sculpts too?

Innula Zenovka

I think it applies to everything. One of the reasons this is being introduced is they want to remove the llVolumeDetect(TRUE) method of phantoming child prims, and have people use PRIM_PHYSICS_SHAPE_NONE instead. Because they don't want to break the huge amount of existing no-mod content that uses the llVolumeDetect trick (which must primarily be sculpts, I think), they're going to make it convert automatically. That's rather worrying, but let's see how it goes.

But they're also aware, going by Falcon's comments in the jira, that under the present accounting system, this would be just as destructive as simply turning off the llVolumeDetect trick, because it would send the prim count of so many converted items through the roof. So they're taking the opportunity to reform the accounting system.

So I think it has to apply to sculpts, since so much of the existing content they're trying not to break is made with them.

Ciaran Laval

Ah so a scripted door in a linkset should no longer cause chaos in terms of LI, that should be helpful.

Qie Niangao

It's telling that this change was motivated by an obscure corner of the backwater that is Pathfinding, as opposed to all the logically-argued mainstream objections that came before.

I guess that is looking a gift horse in the mouth. It's a good development, whatever finally motivated it.

Pandorah Ashdene

Well, I realy hope so - not that I am in any way tempted to go back to making sculpts, - mesh is so much more versatile - but in my experience the LI accounting system as it is now, is the main reason for the slow adaption of mesh as building tool.

Everybody I know, experimenting with mesh, has had their "Ooops" moment, where they linked a mesh element to some kind of 'classical' build (made out of the typical combo of hollowed/cut legacy prims and sculpt elements) and the 'prim count' LI went totaly through the roof by switching to the mesh accounting.
Falcon's example with the three tori is a typical one (- although it's normaly not that extreme with sculpts), where the 'old' LI of 3 jumps to a 'new' 14. Capping the LI at the prim count for legacy content would realy help making the whole building again more predictable and less of a Russian Roulette game.

Ciaran Laval

All we need is a system that doesn't punish us horrendously and encourages us to build efficiently but with an elemnent of fun, the way LI has worked so far hasn't been much fun because of absurd punitive penalties.

Ann Otoole

Stop using scripts and never use avatar attachments and you will meet the standards of LL employees and management. They want SL to suck so it will die.

Been this way since 2007.

And rodvik is a sockpuppet. Bob Komin makes all decisions. Period. Sooner Komin is fired the sooner SL advances.

Drongle

Perhaps I misunderstand, but my reading of the relevant statements is that the change (back) to to one-LI-per-prim will apply only to linksets that are all legacy prims (including sculpties). This is because it is intended only to protect existing content. As soon as there is a mesh involved, the new accounting will still be used on the whole linkset. In that case, the problems of huge LI upon linking meshes to legacy builds will not go away. Someone tell me I am wrong?

Innula Zenovka

I read it as applying to everything. That is, if you take the example of
and then link it to a mesh build, you'll be adding 3, not 14, to the LI.

But I guess someone needs to ask Falcon exactly what it means.

Drongle

Ok, Yes. I read it all again.

"[17:01] Falcon Linden: We're changing streaming cost for prims to be capped at 1.0 and we're changing server weight to be: 0.5 * num_prims + (0.25 * num_scripts) but capped at num_prims"
.... does look as if it applies in all cases, as there is no condition attached. So I can hope I am wrong. On the other hand, the example he gave did not contain any mesh, so it would be good to try to extract an explicit statement that it will still apply when the linkset contains mesh.

If the download weight (formerly streaming cost) is supposed to be just that, then it was indeed very incorrect to weight legacy prims with the LI system, as they are always just the same small parameter set as far as download is concerned. I expect it happened because this became confused with rendering load when it was tied to a per region "triangle budget". Of course the rendering cost of the legacy meshes are high.

If this is indeed a reversion to download resource as the more important factor, that is a good thing. It does bring into question, however, the whole philosophy of using the triangle budget as the basis for scaling download cost, which is currently implicit in the code. Perhaps further changes may be in the air?

Inhandra

The purple's pretty hard to read on dark skin.

__________________

Aribeth Zelin

If he is referring to prim as mesh, built in prims, or sculpts, then its overall a good thing. If he is only referring to built in prims [box, sphere, torus, etc.] than it could be messy.

But I'm reading it as any prim and if so, then it could mean that if you upload a prim that is in parts, that each part is treated as a prim. Since there is an upward limit of amount of faces, and [apparently] using a mesh as a physics model for the mesh does not always give you proper physics, making it necessary to divide meshes into more parts, well, it still might actually help with mesh adoption.

__________________
Faerycat Designs In World -Marketplace 1- Marketplace 2

Ceera Murakami

Well, I hope this is going to be a positive change. I'm a little worried how it will affect things like sculpty stair treads, which often had to use that volume detect hack to make the stairs usable.

So, when does this blow up under our seat cushions? (IE: When does this go live, grid-wide). Or where can we go to see how this will affect some of our existing content?

TaliaDavidov

Yes A question I am very interested in knowing too.

Ann Otoole

it goes into effect when Oz says it will.

Cerise

It will hit the pathfinding test regions first. It will go to the rest of the grid whenever pathfinding is debugged and promoted.

You can find pathfinding regions here. It is supposed to be rolled to those regions this coming week, you can follow PATHBUG-69 for progress.

Ceera Murakami

I went to this region: http://slurl.com/secondlife/Harvey/128/128/25 , and rezzed a set of linked sculpty and normal prim stairs that have a script in the sculpty stair treads that sets volume detect to true for that prim on rez, then disables the script. It was definitely a pathfinding sim - they had bugs crawling on the dock area. But nothing changed about the stairs when I rezzed them there. It didn't have any mesh bits linked to it though. Well, as long as normal prims + sculpted prims in a linkset are not affected by this, I guess I don't care yet. None of my builds have mesh parts in them.

Innula Zenovka

It's not been introduced yet, as far as I know. Falcon was telling people what's going to happen, not what's happening at the moment.

I'm puzzled by your description of how your stairs work, because the volume detect hack only works for unlinked prims that you then link to the main linkset as children. Once they're linked, you can't switch them between pseudo-phantom and not, and haven't been able to for a couple of years. So I think the prims have llVolumeDetect(TRUE) set anyway, and that the script is redundant.

Be that as it may, according to what Falcon says in PATHBUG-69,
So, when the new system is introduced, any changes to your existing builds should happen automatically, but you won't be able to make any more (or unlink and relink the existing ones) without using PRIM_PHYSICS_SHAPE_NONE, but it's not going to have an adverse effect on your build's LI, or that's how I read the comments.

Phoebe Avro

I always get nervous when LL do such things, I may just put my mesh furniture products on hold for a while, don't want a 15 prim sofa becoming 65 prims all of a sudden

Ann Otoole

should be less than 15 prims judging from what i have read so far. As long as there are no scripts. LL does not want scripts anymore.

Innula Zenovka

And even if there are scripts, it's not going to be more than 15.

But -- and this is what I'm really pleased about -- the LI cost of being scripted goes up in proportion to the number of scripts you use, so, while you lose some advantage by putting a script in, you've still got an incentive to be economical.