sl-service-account
commented
5 years ago Vir Linden commented at 2019-07-10T17:59:38Z
This has a lot of discussion relevant to the upcoming Arctan project - will bring it in for review of the various suggestions there.
How would you like the feature to work?
This isn't so much a single feature request but a broader suggestion on how LL could substantially improve the SL experience for all users by reigning in performance killing content.
Drastically changing how Land Impact is calculated, or adding hard limits to the amount of resources individual avatars can use haphazardly would lead to a very upset userbase and likely cause more problems than they resolved, but this doesn't mean the concept of such changes should be dismissed outright. There are ways in which such changes can be implemented while avoiding undesired disruptions for the user base at large.
For example: New Caps on Avatar Resources. Goal: Reduce the amount of high-poly attachments and excessive amounts of texture memory used by individual avatars.
Types of Caps that could achieve this:
The problem is how to introduce such caps without causing the userbase to riot. Here are two methods that could be used individually, or together, to achieve that goal.
When LL switched over from "prim limits" to "Land Impact" they hitched this change to the new feature at the time: Mesh. Mesh content all used to the new Land Impact calculations. If prim content was linked to mesh content, the linkset would all use the new Land Impact calculations instead of prim limits.
Shortly afterwards keyframe animations were introduced. Here again the new rules "Land Impact" were hitched to this new feature. So now mesh and keyframe animations both required users to embrace Land Impact ti take advantage of them. Some time later materials were added to Second Life. We could now use masked alpha textures, normal and spec maps, as well as emissive masks. Here again, any content using prim limits would be switched over to Land Impact if you applied these new features.
I'm suggesting LL could take the same approach to change the rules again.
If we use animesh as the starting point, as an example, any avatar that wanted to wear an animesh attachment would have to accept the new rules as far as attachments go. You could wear new animesh attachments, or detach any animesh to revert to the old rules. This preserves legacy content while giving content creators a push to be more responsible with regards to the content they release. Bakes-on-mesh could add to that. And any new features which come afterwards. If LL released "flexi-mesh" a year or two from now, that too would require people to switch to the new rules to take advantage of the feature.
Over time the number of features would build and more and more SL users would be inclined to adapt to the new rules, at their own pace and without disrupting legacy content. This also means that the benefits of the new rules would not be felt immediately, but I feel LL can afford to take the long view.
This could similarly work for rezzed environmental content in much the same way mesh and materials brought about the change to Land Impact.
A run off effect would be that even those who do not adopt new features right away will still end up acquiring better optimized avatar content as content creators improve their habits.
For landowners, perhaps the new rules could be applied sim or parcel-wide. Switching your land over to the new rules could provide the benefit of a higher Land Impact allotment. After a year or so grace period, new sims could automatically use the new rules giving an additional push for better optimized content.
These examples are not set in stone, they're just meant to illustrate how LL can approach this issue while avoiding any serious disruption for the majority of SL users. While taking these approaches to the problem LL could also apply any new rules to the Aditi test grid. Create a new continent of sims all using the new rules with the associated perks and begin offering land in these sims at a reduced tier rate for early adopters.
All of these options approach the issue in such a way where the SL userbase adapts at their own pace. Old content gets phased out over time naturally, so even if it takes a few years before the improvements are felt widespread, it will happen given enough time.
In addition, perhaps even before LL introduces any new rules, LL needs to engage more with the content creation community. A content creation blog explaining how public content is made, how it's optimized, and how that optimization is reflected in performance would go a long way towards getting content creators to voluntarily change their habits.
Providing users with better tools to manage their own resource use, and to make more educated purchasing decisions, should also be looked into.
I do want to stress that LL does not need to impose new rules that are as strict as, for example, what a professional game designer might hold themselves too. Any new rules LL introduces would only need to curb the worst habits in SL content creators. Limiting avatars to, say, 100MB worth of textures. That's still quite a lot, but it is also far lower than the 300MB many SL avatars tend to be around. That is where the LL content creation blog would come in handy, showing the userbase how to reduce texture use by utilizing more efficient UV layouts and removing unnecessary alpha channels before uploading a texture. Convincing sellers to stop hiding 1024x1024 store logos on hidden faces. This is where the majority of excessive resource use is found in SL content.
Why is this feature important to you? How would it benefit the community?
As things stand, the problem continues to get worse the longer LL fails to act. In a recent interview, Ebbe placed all responsibility for solving this problem on the userbase, a stance which guarantees the problem will only continue to get worse. You have to understand just how much misinformation the userbase believes. Most SL users don't believe excessive texture use has any impact on performance. Content creators won't change their habits if they believe their habits aren't the source of the problems they themselves complain about. On the forum and even in the user group meetings you'll find most SL users are under the impression that the source of all of SL's performance problems are due to Linden error. Content creators avoid any responsibility.
Most of the SL avatars I see these days are using 200-300MB in textures, way up from even just a few years ago. I have seen avatars pushing a full GB just in textures. The problem is getting worse, and will only continue to get worse if LL does nothing. In recent years LL has had to deal with the problem of increasingly frequent crashes due to viewers being forced to handle more memory, this is a direct result and the band-aid fixes applied today won't work tomorrow if things continue as they have.
People tend to greatly underestimate just how much they are affected by badly made content in Second Life. What's more, as LL has remained "hands off" on the issue, it has gotten much, much worse in even just these past 2 or 3 years. Since 2010 I have been attempting to draw awareness to this by crafting a number of high detail sims comparable to the best SL has to offer, while using a fraction of the textures and relying on lower-poly models. Frquently visitors to these sims report double or triple the framerates they typically experience in SL, and almost none of the texture thrashing or performance issues they experience even in far less detailed looking SL environments. To the point where I'm able to run SL on a 13 year old laptop with integrated graphics and still run SL with deferred rendering on and at a respectable framerate.
http://pennycow.blogspot.com/2019/06/just-how-much-does-unoptimized-content.html
This is the kind of performance increase that will help make it easier to draw in potential new SL users who might otherwise be put off by low framerates and excessive stuttering in the current SL experience. SL would be able to run on an even broader range of hardware than it currently does, and frequently at higher graphics settings to boot.
This is not a complete road map, but I hope it can at least prompt a discussion within the Lab on how to take a more active role in addressing the problem. The longer you wait, the more difficult it will be.
Original Jira Fields
| Field | Value | | ------------- | ------------- | | Issue | BUG-227277 | | Summary | Addressing Excessive Resource Use by Second Life Content Creators | | Type | New Feature Request | | Priority | Unset | | Status | Accepted | | Resolution | Accepted | | Reporter | Penny Patton (penny.patton) | | Assignee | Vir Linden (vir.linden) | | Created at | 2019-07-03T13:27:19Z | | Updated at | 2019-07-10T18:00:28Z | ``` { 'Build Id': 'unset', 'Business Unit': ['Platform'], 'Date of First Response': '2019-07-10T12:59:37.849-0500', 'How would you like the feature to work?': 'This isn\'t so much a single feature request but a broader suggestion on how LL could substantially improve the SL experience for all users by reigning in performance killing content.\r\n\r\n Drastically changing how Land Impact is calculated, or adding hard limits to the amount of resources individual avatars can use haphazardly would lead to a very upset userbase and likely cause more problems than they resolved, but this doesn\'t mean the concept of such changes should be dismissed outright. There are ways in which such changes can be implemented while avoiding undesired disruptions for the user base at large.\r\n\r\n For example: New Caps on Avatar Resources.\r\nGoal: Reduce the amount of high-poly attachments and excessive amounts of texture memory used by individual avatars.\r\n\r\nTypes of Caps that could achieve this: \r\n1. A hard cap on total polygon counts of an avatar\'s combined attachments, paired with a hard cap on the amount of texture memory of an avatar\'s combined attachments.\r\n2. Revamping ARC to better account or both texture and polygon use. Currently an avatar can have a very low ARC score, while using excessive texture memory and a high-polygon count.\r\n\r\nThe problem is how to introduce such caps without causing the userbase to riot. Here are two methods that could be used individually, or together, to achieve that goal.\r\n\r\n1. Hitch the new caps to new features. Just in the past year animesh was released, BOM and EEP are set to follow. On top of that, an improvement to animesh is about to go to Project Viewer status. Last year, LL releases a fairly large list of new features and improvements coming to Second Life, so more new features will be coming.\r\n\r\nWhen LL switched over from "prim limits" to "Land Impact" they hitched this change to the new feature at the time: Mesh. Mesh content all used to the new Land Impact calculations. If prim content was linked to mesh content, the linkset would all use the new Land Impact calculations instead of prim limits.\r\n\r\nShortly afterwards keyframe animations were introduced. Here again the new rules "Land Impact" were hitched to this new feature. So now mesh and keyframe animations both required users to embrace Land Impact ti take advantage of them. Some time later materials were added to Second Life. We could now use masked alpha textures, normal and spec maps, as well as emissive masks. Here again, any content using prim limits would be switched over to Land Impact if you applied these new features.\r\n\r\n I\'m suggesting LL could take the same approach to change the rules again.\r\n\r\n If we use animesh as the starting point, as an example, any avatar that wanted to wear an animesh attachment would have to accept the new rules as far as attachments go. You could wear new animesh attachments, or detach any animesh to revert to the old rules. This preserves legacy content while giving content creators a push to be more responsible with regards to the content they release. Bakes-on-mesh could add to that. And any new features which come afterwards. If LL released "flexi-mesh" a year or two from now, that too would require people to switch to the new rules to take advantage of the feature.\r\n\r\n Over time the number of features would build and more and more SL users would be inclined to adapt to the new rules, at their own pace and without disrupting legacy content. This also means that the benefits of the new rules would not be felt immediately, but I feel LL can afford to take the long view.\r\n\r\n This could similarly work for rezzed environmental content in much the same way mesh and materials brought about the change to Land Impact.\r\n\r\n A run off effect would be that even those who do not adopt new features right away will still end up acquiring better optimized avatar content as content creators improve their habits.\r\n\r\n2. Introduce pers SL users can take advantage of under the new rules. For avatars such perks might include the ability to attach an additional animesh object (one would hope that the new rules keeping content better optimized would offset any additional burden one more animesh attachment would add) or an increased number in overall attachments (again, the new rules would keep this from having an adverse affect).\r\n\r\nFor landowners, perhaps the new rules could be applied sim or parcel-wide. Switching your land over to the new rules could provide the benefit of a higher Land Impact allotment. After a year or so grace period, new sims could automatically use the new rules giving an additional push for better optimized content.\r\n\r\n These examples are not set in stone, they\'re just meant to illustrate how LL can approach this issue while avoiding any serious disruption for the majority of SL users. While taking these approaches to the problem LL could also apply any new rules to the Aditi test grid. Create a new continent of sims all using the new rules with the associated perks and begin offering land in these sims at a reduced tier rate for early adopters.\r\n\r\n All of these options approach the issue in such a way where the SL userbase adapts at their own pace. Old content gets phased out over time naturally, so even if it takes a few years before the improvements are felt widespread, it will happen given enough time.\r\n\r\n In addition, perhaps even before LL introduces any new rules, LL needs to engage more with the content creation community. A content creation blog explaining how public content is made, how it\'s optimized, and how that optimization is reflected in performance would go a long way towards getting content creators to voluntarily change their habits.\r\n\r\n Providing users with better tools to manage their own resource use, and to make more educated purchasing decisions, should also be looked into.\r\n', 'Original Reporter': 'Penny Patton (penny.patton)', 'ReOpened Count': 0.0, 'Severity': 'Unset', 'Target Viewer Version': 'viewer-development', 'Why is this feature important to you? How would it benefit the community?': ' As things stand, the problem continues to get worse the longer LL fails to act. In a recent interview, Ebbe placed all responsibility for solving this problem on the userbase, a stance which guarantees the problem will only continue to get worse. You have to understand just how much misinformation the userbase believes. Most SL users don\'t believe excessive texture use has any impact on performance. Content creators won\'t change their habits if they believe their habits aren\'t the source of the problems they themselves complain about. On the forum and even in the user group meetings you\'ll find most SL users are under the impression that the source of all of SL\'s performance problems are due to Linden error. Content creators avoid any responsibility.\r\n\r\nPeople tend to greatly underestimate just how much they are affected by badly made content in Second Life. What\'s more, as LL has remained "hands off" on the issue, it has gotten much, much worse in even just these past 2 or 3 years. Since 2010 I have been attempting to draw awareness to this by crafting a number of high detail sims comparable to the best SL has to offer, while using a fraction of the textures and relying on lower-poly models. Frquently visitors to these sims report double or triple the framerates they typically experience in SL, and almost none of the texture thrashing or performance issues they experience even in far less detailed looking SL environments. To the point where I\'m able to run SL on a 13 year old laptop with integrated graphics and still run SL with deferred rendering on and at a respectable framerate.\r\n\r\nhttp://pennycow.blogspot.com/2019/06/just-how-much-does-unoptimized-content.html\r\n\r\n This is the kind of performance increase that will help make it easier to draw in potential new SL users who might otherwise be put off by low framerates and excessive stuttering in the current SL experience. SL would be able to run on an even broader range of hardware than it currently does, and frequently at higher graphics settings to boot.\r\n\r\n\r\n This is not a complete road map, but I hope it can at least prompt a discussion within the Lab on how to take a more active role in addressing the problem. The longer you wait, the more difficult it will be.', } ```