Distant hills

[animats]

We don't often see distant hills in Second Life. On Strawberry Linden's blog, a recent picture showed some.

North Dakota looks big.

Same location. Ultra, no special settings.

It really does look that big in SL. Go there. It's not done with a huge draw distance. It's one sim, surrounded by off-sim mountains. Those show even outside the draw distance.

Some time back, I was suggesting that SL should do something like this automatically on mainland. Outside the draw distance, draw the bare terrain of distant sims at low resolution. I wondered what that would look like. Now I know. It looks good.

Worth looking into more.

[Scylla Rhiadra]

Wow. Yes, agreed: that looks very good.

[Doc Carling]

Yes, impressing.

[Talligurl]

This reminds me of something that happened to me once. I lived in this apartment on the water. I had this nice ocean view out my window. It wasn't really an ocean though.i was kind of news and didn't know yet about draw distance. I found out about it and so I turned it up. Suddenly the ocean outside my window became a very large river. I could see land on the other side.

[Rowan Amore]

Thanks for sharing, Animats.

[Talligurl]

It is lovely I went there and posted a couple pictures from there in the what do you look like threads.

[Beq Janus]

It's a nice effect, really well done, but in the end it's just old school mega-sculpts used as sim surrounds, it has been done in a more tasteful and visually appealing way than many of the high details surrounds we see,  but apart from the designer's skill there is nothing new here. It still has the same problems, and exemplifies them as a use case.

It is a fantastic example for @Vir Linden, @Ptolemy Linden and Euclid to look at to further the discussion of "We need to solve the sim-surround problem" . I should probably throw all of the following into a Jira and link that to the various Jiras that relate to large scale prims. EDIT: JIRA @ https://jira.secondlife.com/browse/BUG-229551

What we have here is a great example featured on the lab's own blog that relies upon the use of objects that were made in 2006 during a period in which server-side validation was not being performed. The megaprims have long been grumbled about by the lab as undesirable but tolerated anomalies and with recognition that there is no viable replacement. The primary driver is often LI, the prim cap applies, so legacy accounting means that these objects cost just 1LI. However, the example shown is not just cheaper in sculpts, it is not possible without them.

Thus modernising sim surrounds has at least the following problems:-

1. That to do this as a mesh would be implausibly expensive, time consuming and frankly unsightly. 64m tiles even at low resolution. 
   1. The cost is far higher, in this example we have 4x256x256 placed on the corners. giving us 12 64m tiles of "offsim" space. Quick mental math approx is that the LI cost of a triangle at HIGH LOD at 45m radius is 0.06) a sculpt is 2048 triangles for the sake of argument we'll use 1/12 of this per mesh ~170 triangles or 10LI, making it 120LI per sculpt of which there are numerous layers in use in the example.
   2. Complexity in assembly, and use. Due to the linking limitation you cannot link these, you can however described an entire scene in Blender and export it, it will arrive inworld as a collection of unlinked prims. You can then if you are careful manoeuvre it into place, but don't lose the selection or you are in trouble. this is exactly how "Hugh" was imported, and I wrote a special tool to allow me to do both the slicing into 64m hunks and the exporting to DAE as a whole scene for import. Not many creators have that option. Realigning the parts is painful but harder still is ensuring that the seams work properly across LODs...see point 1.3
   3. Unsightly, For all the issues with Sculpt vomit that mesh addresses, aligning the seams of meshes that LOD at different points is a landscaping nightmare. A single 256m object does not have this problem. Once it loads it is done and at that scale it will never LOD swap. The 64m prims will, at normal settings, swap to Medium LOD at just 180m
2. The fact that these render within draw distance is a side-effect of the fact that they are mega prims, the centres of the volumes are thus within draw distance while the geometry is significantly outside of it. In fact, the reality of this is that it is ONLY because of their scale that this can happen at all. prims have to be anchored in the coordinate space of a region. this means that their centre point, or the centre of their linkset parent must be within the regions boinds. The 56m linking distance limitation thus makes it impossible to achieve the level of overlap attained by the creator using sculpts here. Putting aside the LI cost there is no practical way to achieve this effect without sculpts.

If we accept that there are valid uses for these "anomalies" then the case for meshes of similar size and complexity should be considered. They should after all have the same rendering cost as the comparable sculpts, which from the SL documentation is broadly equivalent to a hollow torus prim. 

Using sculpts and megaprims is not, however, a walk in the park either, especially for newcomers.

1. You are limited to the specific range of sizes and shapes that were made before the stable door was bolted shut, in practice this means that you have to have access to the legacy objects, which has got more and more difficult over time (the loss of the SALT HUD was a sad day.) and something that newcomers would frequently struggle with.
2. It also means that you frequently have to use, the closest match rather than an ideal match. It is not as if content creation in SL needs additional learning hurdles to be thrown down. It really doesn't.

Proposal:

A "simple" solution here is to allow meshes to have up to 2048 triangles within 1LI cost irrespective of scale. This would 

What this example does not do is move forward any discussion of truly long-range, low-resolution horizon rendering. It demonstrates how nice things like that might be, it doesn't realistically offer insights into achieving them.

All that aside, It is, undoubtedly, gorgeously put together. Thanks @animats for the heads up.

 

Edited October 24, 2020 by Beq Janus
add a jira link

[Rachel1206]

On 9/9/2020 at 6:37 AM, animats said:

We don't often see distant hills in Second Life. On Strawberry Linden's blog, a recent picture showed some.

North Dakota looks big.

If you are lucky, you will say "Toto, we are not in Kansas anymore", as a tornado hits the corn field!

[Lucia Nightfire]

20 hours ago, Beq Janus said:

A "simple" solution here is to allow meshes to have up to 2048 triangles within 1LI cost irrespective of scale.

Allegedly with ArcTan, mesh scale would no longer affect streaming cost, which would affect land impact.

[Love Zhaoying]

19 hours ago, Beq Janus said:

This would 

..would what?

[Love Zhaoying]

19 hours ago, Beq Janus said:

What this example does not do is move forward any discussion of truly long-range, low-resolution horizon rendering.

Has anyone suggested adding this to EEP? Sky, horizon..same difference.