What the fitmesh LoD bug actually means

[ChinRey]

I finally took the time to do some math.

For those not familiar with the fitmesh LoD bug, it's an error in the code that causes fitted mesh to be rendered at much higher complexity than necessary. That's bad enough but to make matters worse, the formula used to calculate render complexity, does not know about this bug and does not take it into account.

Lots of tricky stuff below so to save people the effort, here are the results:

• A typical fitmesh hair, uploaded at aproximately realistic size, a 1024x1024 texture, triangle counts for the LoD models 40,000, 1,000, 100 and 2 respectively:
  • Calculated render complexity: 29,563
  • Actual render complexity: 144,512 - about five times as high as the calculated one
• The same hair uploaded at minimum size to take full advantage iof the cheating possibiities of the LoD bug:
  • Calculated render complexity: 520
  • Actual render complexity: 144,512 - about 280 times as high as the calculated one

---

Now for the calculations. If somebody spots any errors here, please correct them.

The official formula for calculating render complexity is here: http://wiki.secondlife.com/wiki/Mesh/Rendering_weight. That formula has three problems though:

• The multiplier used for calculating base weight in that formula is not correct though, the actual one is much lower than 5. It's hard to say exactly what it is but it seems to be somewhere around 3 and I've used that value for my calculations. (I've added some info how I got to that number at the end of the post). This does not affect the relative render cost size between my three examples very much though and that's what matters here.
• The formula does not mention fitted mesh at all. I assume it uses the same 1.2 multiplier as rigged mesh but that's just an educated guess. Again, this does not affect the relative render cost since whatever the multiplier is, it applies to all three calculations.
• The formula is old and there may well be undocumented changes. It's the only one we have though.

The formulas I used for calculating LoD swap distances are:

• M=(√(x*x+y*y+z*z))/0.6
• L=M*4
• Lw=M*8
• C=181

 

• x, y and z are the object's dimensions along the axises
• M is the swap point from high to mid LoD
• L is the swap point from mid to low
• Lw is the swap point from low to lowest
• C is the clamp distance

---

Take a fitted mesh with 40,000 triangles in the high LoD model, 100 in the mid model, 10 in the low and 2 in the lowest. Texture it with a 1024x1024 pixel texture. Wear it on a mesh avatar with bloated bounding box the way Beq Janus described it.

The average number of triangles displayed across the view distance radius is 40,000 since this mesh will always be render at high LoD. That gives us a render cost of 144,512.

But the render cost formula doesn't know about the fitmesh bug. It assumes the LoD system works as it was suppsoed to and bases the LoD swap distances on the size the mesh was uploaded at. Let's say that size was 0.5x0.5x0.5 m - about right for a mesh hair. That means the forumala believes the average number of triangles is only 6,917, giving a calculated render cost of 29,563, only a fifth of the actual cost.

It gets worse. Many fitmesh makers know about the LoD bug now and they know how to take full advantage of it by uploading at the smallest possible size, making their meshes seem far less laggy than they actually are. (Other makers do this without even knowing about it since at least one mesh program creates undersized meshes by default.) In that case the calculated average number of triangles becomes 2 (since only the lowest LoD model really counts) and the calculated render weight ends up at only 520.

---

Finally, the base weight multiplier:

• A 64x64x64 m prim torus with a 32x32 pixel texture should have a render weight of 5384 according to the official formula. The calculated number is 3749.
• A prim cube with the same size and texture should have 804 render weight. The number given is 269.
• A mesh with 392 triangles in all models and a 32x32 pixel texture should have 224 render weight. One I uploaded had 1854.

It's hard to draw any firm conclusions from this since the viewer doesn't count the number of triangles when it calculates render cost. It just estimates the number from the file size.

Edited August 22, 2018 by ChinRey
Correcting typos

[Coffee Pancake]

The numbers are so far away from the posted math that it's more than likely they are simply long out of date.

T' lab doa change stuff 'n dooant update t' wiki.

[ChinRey]

40 minutes ago, CoffeeDujour said:

The numbers are so far away from the posted math that it's more than likely they are simply long out of date.

I'm afraid not.

That is, the formula may have been tweaked a bit over the years (maybe somebody from LL or somebody who knows the viewer code can answer that) but i hasn't been drastically changed. And any modifications would have had pretty much the same effect on all three calculations so the relative difference between them wouldn't change much.

So no, if the render complexity formula had been applied to fitted mesh the way it was intended and should have been done, with the correct LoD swap distances, the render cost of a fitted mesh would have showed up with a render cost several times higher than it does now with the borked calculation and a mesh that exploits the bug might well have showed up with a render cost several hundred times higher if the formula had been used correctly.

I know it's extremely unpopular to say this. Nobody really wants to know how laggy our mesh avatars are - I sure don't. But facts are facts.

We can of course discuss how reliable the render complexity formula is. It has several other flaws too. But none of them seem to be anywhere near as serious as this.

[Fionalein]

So basically when I get booted from an event for flexi hair and return with mesh hair I am rewarded by being allowed to stay for making it worse? Huzzah!

Edited August 22, 2018 by Fionalein

[Coffee Pancake]

30 minutes ago, ChinRey said:

I know it's extremely unpopular to say this. Nobody really wants to know how laggy our mesh avatars are - I sure don't. But facts are facts.

No argument here. Our current avatars are unacceptably heavy, but nothing will change unless the Lab make them impractical for day to day use and the BRAND_NAME vendors release a low impact version intended for daily use ... I mean, this would be an awesome thing for them to do anyway.

There is no reason why you can't have a superdetail body, but you shouldn't be taking it to the shops or a club.

[ChinRey]

1 hour ago, CoffeeDujour said:

The numbers are so far away from the posted math that it's more than likely they are simply long out of date.

Sorry about double posting but it occured to me that it's a fairly easy reality check we can do.

Regardless of any changes that may have been made to the forumla, what we do know, is that when a fitted mesh is uploaded with a size of 0.1 m or less along all axises, the render weight formula will assume that the mesh will nearly always be rendered at lowest LoD and that the other models are so rarely used they don't matter to the base weight. In fact, if it's fitted mesh, it will nearly always be render at the highest LoD.

We can test what difference that makes with regular meshes where the LoD bug doesn't apply. Here are two of them. The one to the left is just a plane with two triangles, what the RC formula assumes the fitted mesh in my third calculation is rendered as. The cube to the right has been split into 40,368 triangles, what my fitted mesh example actually is rendered as. Both meshes are textured with the same 1024x1024 texture.

The two triangle mesh has its render weight calculated to 517:

The one with 40,368 triangles is 100,822

 

[ChinRey]

2 minutes ago, CoffeeDujour said:

There is no reason why you can't have a superdetail body, but you shouldn't be taking it to the shops or a club.

Yes but one problem is of course: how do you know how laggy your avatar is? The render complexity system is so compromised by this bug alone it's completely useless and that's the only thing we have.

[Coffee Pancake]

3 minutes ago, ChinRey said:

Yes but one problem is of course: how do you know how laggy your avatar is? The render complexity system is so compromised by this bug alone it's completely useless and that's the only thing we have.

You shouldn't need to care. Which puts the ball squarely in the hands of content creators and they will only respond to hard limits.

We have hard limits for land in the form of Li. We should have those for avatars too ... getting more would be a decent premium perk too.

[Fionalein]

14 minutes ago, CoffeeDujour said:

We have hard limits for land in the form of Li. We should have those for avatars too ... getting more would be a decent premium perk too.

@Klytynawill like this idea... 

[Callum Meriman]

4 hours ago, CoffeeDujour said:

We have hard limits for land in the form of Li. We should have those for avatars too

/me washes your mouth out with soap then sends you to the naughty corner.

No, we should not have hard limits for avatars. An absolutely shameful idea that dictates how I may dress in my own region with nobody but my partner with me.

[ChinRey]

8 minutes ago, Callum Meriman said:

No, we should not have hard limits for avatars. An absolutely shameful idea that dictates how I may dress in my own region with nobody but my partner with me.

It could be similar to the land impact limit: A limit not to the load of each avatar but to the total load of avatars in the region.

But before we can even consider something like that, we need a reasonably reliable way to estimate the load an avatar causes.

[Callum Meriman]

Just now, ChinRey said:

It could be similar to the land impact limit: A limit not to the load of each avatar but to the total load of avatars in the region.

Disgusting Idea! You can go to the naughty corner too!

You are trying to penalise people for something that jellybeans already solve.

If people run on potatos, then they can set their maximum complexity to a level that gives them good FPS, and in addition they can manually derender people.

Calling for a hard limit is braindead stupid

 

 

[ChinRey]

1 minute ago, Callum Meriman said:

You are trying to penalise people for something that jellybeans already solve.

Well, as I said, first lets get a way to measure the render load of avatars, then we can see what measures are needed to control it.

Jellybeans does not solve anything at all now because the numbers it is based on is waaaaaaaaaaaaaayyy off. As I've said before, you might as well throw a dice to determine which avatars to render fully and which to render as jellydolls as trust what the viewer decides. But if the software is fixed, it may be enough. We'll see.

[Callum Meriman]

3 minutes ago, ChinRey said:

then we can see what measures are needed to control it.

We do NOT need a restriction on how complex an avatar can be, with premiums allowed more.

We do not need a limit that says a homestead may only have 600K complexity before people are blocked from entry.

We already have the solution to complex avatars - if you run a potato you can set your jellybean threshhold lower. You can also derender people with prim tortured hair, flexis, and jewellry.

There is NO need to impose complexity limits on avatars.

Grrrrr.

[ChinRey]

1 minute ago, Callum Meriman said:

We do NOT need a restriction on how complex an avatar can be, with premiums allowed more.

We do not need a limit that says a homestead may only have 600K complexity before people are blocked from entry.

We already have the solution to complex avatars - if you run a potato you can set your jellybean threshhold lower. You can also derender people with prim tortured hair, flexis, and jewellry.

There is NO need to impose complexity limits on avatars.

Grrrrr.

Chin Rey sighs.

Callum, let me spell this out as simple as possible:

We do not at the moment have any way to measure or control avatar complexity. The whole quick graphics/render complexity/jellybean system is nothing but a poor joke because there is no way for the viewer to know how laggy an avatar actually is. That is what this thread about.

As it is now, we have no way of knowing which avatars are the laggiest ones and until we know that, there is no point in even trying to limit avatar render cost. We won't have that soon and we may well never have it. So let's just put that discussion aside for now. Any reasonably effective solution to control avatar render cost is so far into the future there's no point discussing it now.

[Callum Meriman]

18 minutes ago, ChinRey said:

So let's just put that discussion aside for now.

No, this discussion about "hard complexity limits" needs to be put aside forever. Coffee and you are very wrong to raise it.

The purpose of complexity, of ARC is to provide people with the tools they need to adjust the game to suit their computers. It's not there, and should never be there to provide a hard limit on what someone may or may not wear on their avatar. Especially with braindead ideas such as premiums can wear more stuff. That's unfair, and it's stupid because people already have a slider they can lower to turn people into jelly beans. They already have the ability to derender people they judge as unimportant on top of that.

And there is absolutely no need to set a new entry limit on homesteads so that people simply can't enter the region when it's reached a total avatar complexity of 600K. That's unfair again on people who have decent computers, and it's unfair on region - especially club - owners.

This is the second time we have seen this braindead suggestion that people need to be allocated a hard limit on what they may wear, and this needs to be the final and last time we see it.

You have no right to tell me what I am allowed wear in my regions. Stop asking for this to happen.

[ChinRey]

1 hour ago, Callum Meriman said:

No, this discussion about "hard complexity limits" needs to be put aside forever. Coffee and you are very wrong to raise it.

Ok. Consider it done.

 

1 hour ago, Callum Meriman said:

The purpose of complexity, of ARC is to provide people with the tools they need to adjust the game to suit their computers.

Yes and it would have been really nice to have such a tool. Sadly, we don't.

[Coffee Pancake]

2 hours ago, Callum Meriman said:

You have no right to tell me what I am allowed wear in my regions. Stop asking for this to happen.

You are already limited to a set number of attachments and will be limited to a single animesh attachment. Back when prims were a thing, you were limited as to how many could be combined (which baked in a upper limit on possible complexity). There have always been limits. Mesh attachments on the other hand ... not so much.

Jelly dolls do not solve the problem when the measuring stick we have to guage avatar rendering weight is being gamed to the point of uselessness. The only avatars who get jellys are those who don't buy hacked in stuff. Did you even read the OP ?

There is no way for people to dynamically adjust the complexity of avatars, it's render them fully, or don't. On or off. Maybe if the LOD system wasn't being abused and junk that would be a thing, but it's not. A social platform that removes people in answer to a technical problem has a fundamentally problem.

What we need is for mesh body and clothing makers to take the rendering impact of their products as a serious concern and make content that is fit for the platform. They are not doing this and if anything has been proven true about SL trends, it's that more detail sells. 

There are no guidelines, no benchmarks, nothing to aim for. The workflow is to make the thing and finger in the wind figure it's good enough, then work out how you can get it into SL for another finger in the wind Li/upload cost

 

ARC provided a way to guage content. The response of the creative industries was not to make content with ARC in mind, it was to find a way to ram terrible content and cheat.

 

So what is the solution? Carry on .. 

In case you hadn't noticed SL is CPU bound and depends entirely on it's straight line single core speed. CPU's aren't getting faster, they are getting wider and that doesn't help us one bit.

Fancy new graphics card sitting at 40% use when running SL and the framerate stinks, because the CPU can't go any faster.

We're sitting in a hole and it's only getting deeper. The numbers are dropping as people on the low end give up. It might shock you .. but the vast majority of the SL user base are on the low end running SL on laptops.

 

Edited August 23, 2018 by CoffeeDujour

[Callum Meriman]

19 minutes ago, CoffeeDujour said:

Jelly dolls do not solve the problem when the measuring stick we have to guage avatar rendering weight is being gamed to the point of uselessness. The only avatars who get jellys are those who don't buy hacked in stuff. Did you even read the OP ?

Yes, I read the Op and I agree with it. I hope there is a jira for it so that it can be taken into account for project Arctan.

What I do not agree to is your post saying:

8 hours ago, CoffeeDujour said:

We have hard limits for land in the form of Li. We should have those for avatars too ... getting more would be a decent premium perk too.

We should NOT have hard limits for avatars. What we choose to wear in our regions, in clubs we own and pay for is our business.

If events wish to place a complexity cap - as some do, then they can - but for you to come out and recommend that everyone in SL should abide by YOUR idea of what a person should wear is wrong, and it's wrong to the core.

To make it worse, you actually request that premiums should be allowed to wear more complex avatars then basics.

"Sorry Ma'am, you cant have that hair - it's only for premium members"

 

 

 

[ChinRey]

1 hour ago, Callum Meriman said:

Yes, I read the Op and I agree with it. I hope there is a jira for it so that it can be taken into account for project Arctan.

I don't know if there actually is a JIRA but Vir Linden has confirmed that they are aware of the bug and I got the impression they were looking at a solution as part of project Arctan.

The question is what are they going to do about it? I can only see four possible answers to that:

---

1. Correct the calculation to reflect the actual render cost of fitted mesh

That is probably the easiest solution but it has some serious consequences.

One is that the render weight of mesh avatars will go through the roof. The jellybean filter has a maximum cutoff of 300,000. Most mesh avatars will probably be far higher than that. The lsl function is capped at 500,000. Some mesh avatars will turn out to have render weights in the millions.

The second consequence is that people will feel cheated when they learn how laggy their avatars actually are. There's going to be an outrage.

The worst consequence, however, is that it will cement the current state with no LoD simplification for fitted mesh at all. I think that will be disastrous in the long run.

---

2. Fix the LoD system so it works as itnended for fitted mesh

I think that's the only real solution in the long run. But I've looked at several fitted mesh items of different kind and I've yet to see one that would survive the LoD system. Expect every single fitted mesh in your inventory to be broken if they go for that solution.

---

3. Introduce an automated LoD system for fitted mesh

That's a huge project, far bigger than Arctan. I can't imagine LL has the capacity to do that.

---

4. Out some bandaid on the wound and pretend it's ok

Maybe a symbolic little increase in the render weight just to show that they take it seriously. Well, any of the three first options would require far more integrity, dedication and - above all - courage than LL has ever shown so I strongly suspect this is what they're going for.

Edited August 23, 2018 by ChinRey

[OptimoMaximo]

4 hours ago, Callum Meriman said:

No, this discussion about "hard complexity limits" needs to be put aside forever. Coffee and you are very wrong to raise it.

Sorry to contradict your verdict, your Highness (because you must be high on something to claim no limits), but there is a desperate need for such control and YOU are very wrong in your claim for no limits, this discussion MUST be raised over and over and over again until LL finally listens and takes action. Real action, not the jokes you refer too as "solutions" like the jellydoll. There are standards that MUST be complied with when it boils down to realtime 3d content, exactly like any other platforms.

 

4 hours ago, Callum Meriman said:

You have no right to tell me what I am allowed wear in my regions. Stop asking for this to happen.

The problem is that regions do not have anything to do with what you wear, it's the VIEWER that handles attachment rendering. And if you stand in a region while you're within my view range when i sit in MY region, and YOU bring up my GPU temperature over 80 Celsius (and it's not a low end one), YOU are the problem, not my GPU. So please, YOU stop claiming no avatar complexity limits/control. There are brands that sell MOVIE standards meshes into SL, when those meshes are being used for final renders on renderfarms and a lower poly version of them is actually used in production for PERFORMANCE, ON STANDALONE APPLICATIONS.

The use of actual LoD systems, on the other hand, can help raising these limits, but only if the LoD switch works. As it stands now, it doesn't and needs a fix. So let's be serious, you're not the only one in SL and other people machines and needs must be taken into account.

1 hour ago, ChinRey said:

1. Correct the calculation to reflect the actual render cost of fitted mesh

That is probably the easiest solution but it has some serious consequences.

One is that the render weight of mesh avatars will go through the roof. The jellybean filter has a maximum cutoff of 300,000. Most mesh avatars will probably be far higher than that. The lsl function is capped at 500,000. Some mesh avatars will turn out to have render weights in the millions.

The second consequence is that people will feel cheated when they learn how laggy their avatars actually are. There's going to be an outrage.

The worst consequence, however, is that it will cement the current state with no LoD simplification for fitted mesh at all. I think that will be disastrous in the long run.

---

2. Fix the LoD system so it works as itnended for fitted mesh

I think that's the only real solution in the long run. But I've looked at several fitted mesh items of different kind and I've yet to see one that would survive the LoD system. Expect every single fitted mesh in your inventory to be broken if they go for that solution.

These two points together are the solution, not either/or. In RL, when a business cheats on the customer and is caught in cheating, there are consequences to face (or cope). The System can't allow the cheating to stay and continue unpunished. At least in my State, when you buy a counterfeit good and are caught buying it, even if you didn't know it wasn't original, you get a fine along with the seller. The seller gets a worse treatment, higher penalties and a lawsuit, of course. The lazy cheaters need to run out of business or line up to the guidelines with no cheating. This results in hundreds if not thousand updates to existing content? None of the System problem, considering that is a sort of "reward" for those who didn't cheat in the first place. Everyone is able to grab or make a mesh, detail its geometry to death and slap a lazy texture on it just because the geometry does the work. This is a game-like environment, not a feature film, and content must be made according to game requirements. Takes more time and can't produce as fast as the other method? Awww so sorry...

1 hour ago, ChinRey said:

3. Introduce an automated LoD system for fitted mesh

That's a huge project, far bigger than Arctan. I can't imagine LL has the capacity to do that.

This is sadly out of question. Stored assets can't be manipulated in their architecture, it should be a thing for new uploads (and a new upload system which enforces such a thing), if this ever is going to be considered by LL.

1 hour ago, ChinRey said:

4. Out some bandaid on the wound and pretend it's ok

Maybe a symbolic little increase in the render weight just to show that they take it seriously. Well, any of the three first options would require far more integrity, dedication and - above all - courage than LL has ever shown so I strongly suspect this is what they're going for.

And i hope this is not what is going to happen, although i'm afraid you're right. You're right by saying it's a matter of integrity and it should be enforced somehow, although not as harshly as a State would act. A "grace time" to update products, before the fixes go live, to all the "cheaters" without the obvious naming and shaming and outrage that would go with the sudden breakage of their content. Who has cheated knows what they did, fortunate enough to get some time to right your wrongs... who doesn't know about it, instead, has to face their incompetence and their customers need to realize who they were trusting as a "pro".

3 hours ago, CoffeeDujour said:

We're sitting in a hole and it's only getting deeper. The numbers are dropping as people on the low end give up. It might shock you .. but the vast majority of the SL user base are on the low end running SL on laptops.

Because of this, indeed, LL had started Project Sansar. The user base with low end machines is dropping SL, and the highpoly = highquality trend is the cause of it, which in the long run is going to shatter the economy into dust with consequent platform shut down. Is cashing out as fast as possible now worth the lack of cash outs for years to come, considering the ridiculous development of Project Joke Sansar? If it has to be considered business, as a business it should be treated and try to keep it up as long as it is possible to.

[Fionalein]

The bad thing is: folks have unreplaceable no copy items made from many tortured prims... Who is going to tell Mrs. Smith her engagement ring cannot be worn anymore because LL said so? The outrage would be devastating - it is one thing to ask Mrs. Smith to remove the ring for an event - difficult enough. Telling her she can now dump it will cause uproar.

Edited August 23, 2018 by Fionalein

[steeljane42]

4 hours ago, OptimoMaximo said:

The user base with low end machines is dropping SL

I'm actually not sure if LL care or should care. Such low end users usually are not a part of the SL economy anyway, aside of in-world jobs and if they are, money they earn are still coming from paying residents. A used hardware within 500$ range (cpu/mb from sandy/ivy bridge era with overlock and 30$ cooler, gtx 960, other basic parts are going to be "fine") can provide a decent SL experience. Decent is not perfect and I don't argue that it shouldn't be improved, but not via such radical methos, like enforcing avatar complexity limits on everyone.

So shall LL cater to 100% F2P userbase (and a few "fighers for lag-free SL" from this thread) or paying customers? I'd say it's an obvious choice. Fix bugs, improve things, but never ever limit your paying customers on what they already got used to through the years. Mass exodus of people who can't use what they bought anymore vs possible (yes, possible, SL has more issues than just bad performance) fresh flow of users with extremely low end machines, who are most likely won't spend anything on SL. That's how I see it anyway.

So yeah, I'm all of LL fixing this LoD bug and rework jellydol system so it actually wouldn't render such jellied avis, as long as it doesn't affect how I see my friends and they see me and what stuff I can use. Low end users can't see me already anyway, so if fixing this bug will make my complexity from 50-150k to 1m... it won't really change a single thing for me, it's just numbers.

[ChinRey]

1 hour ago, Fionalein said:

The bad thing is: folks have unreplaceable no copy items made from many tortured prims...

Oh, those shouldn't be a problem. The LoD system does work as intended for prims, sculpts and rigid mesh so unless it's made from fitted mesh, something as small as a ring will always be rendered at lowest LoD except when you cam in really, really close. The torus is the prim with the highest polycount and it still only has 72 triangles at lowest LoD . Torturing it can often reduce the polycount further and never ever increase it.

 

4 hours ago, OptimoMaximo said:

In RL, when a business cheats on the customer and is caught in cheating, there are consequences to face (or cope).

Yes but this is really Linden Lab's responsibility. It's not even always deliberate cheating. Export a fitted mesh from Maya to SL and it ends up at minimum size whether you intended it or not. At least that's what you told me earlier.

 

5 minutes ago, steeljane42 said:

Such low end users usually are not a part of the SL economy anyway

Really? Linden Lab disagrees with you. They are trying their best to keep their official builds as low lag as possible because so many premium members are on low end computers. That's official and I do actually believe they know best in this case.

An old-timer builder friend of mine told me LL had tried to persuade him to try Sansar. He didn't dare tell them he doesn't have a computer strong enough to handle Sansar. He pays Linden Lab tier for half a sim for his store.

Me? I'm not in the danger zone yet but if the lag infaltion goes on, I won't be here two years from now. If that happens, that'll be 500 dollars a month lost for LL.

Those are just a few examples. All indications I've ever heard or seen indicate that what you call "low end users" contribute more to the SL economy than the few with expensive hardware.

[OptimoMaximo]

8 minutes ago, steeljane42 said:

So yeah, I'm all of LL fixing this LoD bug and rework jellydol system so it actually wouldn't render such jellied avis, as long as it doesn't affect how I see my friends and they see me and what stuff I can use.

Since you quote just a section of what i said for your convenience, I'll do the same to my convenience.

Aside from the fact that the "stuff you can use" shouldn't be here in the first place, but it is because of lazy cheating, the paying customer you claim so important usually started 1) as new user 2) with F2P intentions because SL is advertised as such 3) on low end machines that got upgraded overtime. On this latter, how would LL retain a new user if "the game melted my GPU after few log ins"?

Then again, you want jelly dolls to go away AND the fixes, but ONLY AS LONG as YOU're not affected. Reality check: the Universe doesn't revolve around you. Guess what? After the LoD fixes we hope, with your current stuff, you'll cam 1 centimeter away from your pretty face and your meshes will crumble into single triangles for each piece it is made of. So sorry, go ask the creator why it happens, but i doubt they will ever answer to you saying the truth, which simply is "i had to pose as a pro, cheating on the system, to make my stuff look like high quality but i couldn't get it to work with a proper LoD system because i'm an illiterate on the matter, and as long as there is kettle to be milked as quick and as much as possible, like you, i couldn't care less". Second reality check: high resolution meshes do not belong to this kind of environment, and if you get them in, it's to be considered CRAP. Regardless whether it looks good. Also because what you pay isn't consisting only in the mesh itself, add your video card reduced life span to it. Those creators are the source of the problems, not LL. And as such, once LL cleans up the mess, those creators have to fix the problems you get from the clean up. I don't know about you, but usually people want to live in a clean place, not among piles of crap.

Then you talk about "improvements"... so reducing lag (which is not the goal of such discussion, anyway, but apparently it's the only thing you are able to get about it all) isn't among the things anyone would include in the list of improvements, along with machine's rendering overhead that reduces other performances you aren't even aware of. Right. Well said. Being "just numbers", as you said, will definitely affect how you will see others and how others will see you... you know, this and all virtual worlds RUN ON AND BECAUSE OF NUMBERS.