Linden Lab lays off 30 staff

[Chalice Yao]

I really wouldn't read too much into the laying off of the 30 Sansar employees - at least nothing in regards to another big player buying them, planning to open up a rival feeling threatened by Sansar, and such things, etc.

I checked yesterday, and according to Steam stats, Sansar there still has around 30-40 concurrent users total on average. A sim's worth.

Given how much money LL has spent on Sansar, how much time has passed since launch, and how little the stats have improved - especially compared to things like VRChat , which has around 6k to 7k concurrent users now, and got those in a heartbeat - anybody seeing Sansar of all things as a 'rival' would be kind of amusing.

It's much more probable that LL is - finally - officially deeming Sansar a failure. With those...cute...user stats two and a half years after Sansar opened its doors to the public, it hardly can be treated as anything else. It was a huge, huge blunder.

 

[Chav Paderborn]

Does anyone know what SL's concurrency was like at a similar age?

 

[Bartholomew Gallacher]

Even the most advanced triple-A games designed to be played on cutting-edge consumer computers still realize the need to use simplified meshes with normal maps or bump maps on them that provide the real "detail". The way that whole system works is, you first need to build a fantastically dense and complex but -completely unusable for purpose- mesh model, which exists solely to make a normal map out of for you to apply to a second, much simpler mesh model which is the only one of the two models you can actually put into your game (i.e., upload to SL).

Meshes for AAA games are designed by highly paid professionals, who exactly do know about how much complexity they could use and the dirty little tricks of the used engine inside out - so this is no wonder. SL meshes on the other hand most of the times are not.

 

[Bartholomew Gallacher]

I thought both competed with each other for a while.

No they did not really. There was a JIRA ticket, where Maxwell Graf and others showed up that a mesh deformer might be beneficial, and it was downvoted. Considering the stellar track record Linden Lab's in being a snail they then fired up an IndieGogo campaign and started coding on their own.

Mesh Jira SH-2374 downgraded to maybe someday

Well, this isnt very good news. Apparently, the Jira I posted about mesh clothing, has been downgraded from Sprint 28 release to "someday/m...

rusticahomefurnishings.blogspot.com

When Linden Lab then decided to implement their own solution the coexistence was over.

 

[Chalice Yao]

Meshes for AAA games are designed by highly paid professionals, who exactly do know about how much complexity they could use - so this is no wonder. SL meshes on the other hand most of the times are not.

Eh, not sure about "highly paid" these days.

But yeah. The majority meshes you see at the big SL Mesh and Designer events would not, at all, fly in any modern game production. Not even just AAA games, but more often than not they fail all sorts of things that are standard practice and expected skills in modern game design:

* Efficient geometry. Most popular SL creations have way, way too dense geometry. And it's reached and spread inside the hair market as well by now, sadly.
* Efficient UV layout. Dear god, this one is SUCH A HUGE disaster in SO MANY CASES. So many creators just auto-.generate some UV that is mostly blank space, bake the textures in, and don't even *start* working on the UV by hand. At all. And it worsens the next problem by a mile.
* Texture usage. The crappy UV layouts make it so that what could be a 256x256 (or even a 128x128) needs a 1024x1024 to reach the same visual quality. And A 1024x1024 on an outfit's single button texture would get you laughed out of the building when showing off your oh so 'professional' SL portfolio at a game company.

Many of the high-tier SL creators like to promote themselves as if they are professional and that their creations are high-quality, but the fact is that the only market for the type of meshes they are making is here, in Second Life.
Too high geometry and texture inefficient for any game's designer position and game asset usage, too low quality for anything CG.
The only market for them is SL where people don't know better, and hardly have any other options these days.

 

[Chin Rey]

I've always thought that Sims 4 style home building tools would be wonderful in SL. I'm not sure how doable it would be, but it seems like it would open up a lot of new possibilities, including third parties being able to make parts that would work with the system and be added to the user's library of available things to build with.

We already have something sort'a like that with full perm prefabs.

 

[Fionalein]

Eh, not sure about "highly paid" these days.

But yeah. The majority meshes you see at the big SL Mesh and Designer events would not, at all, fly in any modern game production. Not even just AAA games, but more often than not they fail all sorts of things that are standard practice and expected skills in modern game design:

* Efficient geometry. Most popular SL creations have way, way too dense geometry. And it's reached and spread inside the hair market as well by now, sadly.
* Efficient UV layout. Dear god, this one is SUCH A HUGE disaster in SO MANY CASES. So many creators just auto-.generate some UV that is mostly blank space, bake the textures in, and don't even *start* working on the UV by hand. At all. And it worsens the next problem by a mile.
* Texture usage. The crappy UV layouts make it so that what could be a 256x256 (or even a 128x128) needs a 1024x1024 to reach the same visual quality. And A 1024x1024 on an outfit's single button texture would get you laughed out of the building when showing off your oh so 'professional' SL portfolio at a game company.

Many of the high-tier SL creators like to promote themselves as if they are professional and that their creations are high-quality, but the fact is that the only market for the type of meshes they are making is here, in Second Life.
Too high geometry and texture inefficient for any game's designer position and game asset usage, too low quality for anything CG.
The only market for them is SL where people don't know better, and hardly have any other options these days.

to be fair 3d movie studios would not mind higher details - if justifiied that is ... render farms still cost but in a movie the end justifies the means if higher rezolution will benefit the movie they will more likely tolerate it.

 

[Chalice Yao]

to be fair 3d movie studios would not mind higher details - if justifiied that is ... render farms still cost but in a movie the end justifies the means if higher rezolution will benefit the movie they will more likely tolerate it.

Yeah, that's what I meant with 'Too low quality for anything CG' - depending on the production, SL's asset - even many of the really unoptimized ones - are *way* too low poly in geometry and especially textures.

 

[Argent Stonecutter]

HAHAHA you think so? Well we're talking about an imaginary SL feature that never existed here so I guess I can't really prove you're wrong about that. I will simply point out that both Second Life's and Sansar's user interfaces have never been upheld as shining examples of UI design, not as long as I've been a user at least.

The original SL UI was not bad. SL2 is the one that went over the edge, but even in SL2 the object editor is way way better than anything in Blender.

Does anyone know what SL's concurrency was like at a similar age?

I got into SL in mid-2005, and it had opened in 2003, and I don't know the 2005 concurrency figures but even back then the average telehub had more avatars in it than all of Sansar today.

 

[Essence Lumin]

Does anyone know what SL's concurrency was like at a similar age?

I can't give you a number but can tell you that in October 2003 there were considerably more than 20 or 30 people online at a time.

 

[Bartholomew Gallacher]

Tattere Nino in former times had a bunch of statistics - but it seems no archive.

Tyche Shepherd then again offers the daily concurrency count via API - so probably has an archive. But Tyche joined in 2007, so quite sure nothing before that time.

 

[Chalice Yao]

but even in SL2 the object editor is way way better than anything in Blender.

Yeeeeah.....noooo, not anymore.

 

[Daniel Voyager]

I can't give you a number but can tell you that in October 2003 there were considerably more than 20 or 30 people online at a time.

Second Life User Concurrency Peaks

• 88, 220 - 29th March 2009 at 1:28pm SLT
• 86, 584 - 15th March 2009 at 3:57pm SLT
• 81, 913 - 28th February 2009 at 2:50pm SLT
• 80,000 - 12th January 2009
• 80, 528 - 11th January 2009 at 2:46pm SLT
• 80, 195 - 11th January 2009
• 74, 990 - 21 December 2008 at 2:04pm SLT
• 71, 232 - 28th September 2008
• 70, 821 - 21st September 2008 at 1:58pm SLT
• 69, 574 - 14th September 2008
• 66, 429 - 30th March 2008
• 60, 000 - 14th January 2008 at 1:45pm SLT
• 58, 738 - 6th January 2008 at 1:40pm SLT
• 56,967- 5th November 2007 at 1:50pm SLT
• 51, 914 - 2nd November 2007 at 2:25pm SLT
• 50, 872 - (min 27,105) - 31st October 2007 at 3:30pm SLT
• 50, 000 - 4th September 2007
• 42, 285 - 6th May 2007
• 39, 001 - (min 20,074) - 1st May 2007 at 1:31pm SLT
• 31, 000 - 2nd February 2007
• 20, 015 - Mid December 2006
• 18, 000 - Early December 2006
• 14, 000 then by end 16,000 - November 2006
• 11, 000 - 1st October 2006
• 10, 000 - 27th August 2006
• 9, 613 - 30th July 2006
• 7. 005 - 6th June 2006
• 6,000 - March 2006
• 2. 127 - June 2005
• 1, 466 - 13th May 2005
• 666 - 18th March 2005

SL Unique Residents Population

• 2003 April 227
• 2003 May 338
• 2003 June 527
• 2003 July 758
• 2003 August 924
• 2003 September 1,088

Second Life Signups: 2005-2015

• 28, 471 - 13th May 2005
• 30, 014 - 30th May 2005
• 40, 007 - 3rd August 2005
• 100, 000 registered avatars - 23rd December 2005
• 200, 000 registered avatars - 28th April 2006
• 300, 000 registered avatars - 29th June 2006
• 400, 000 registered avatars - Early August 2006
• 500, 000 registered avatars - Later in August 2006
• 800, 000 registered avatars - October 2006
• 1 million registered accounts: 18th October 2006
• 2 million registered accounts: 14th December 2006
• 3 million registered accounts: 28th January 2007
• 4 million registered accounts: 24th February 2007
• 5 million registered accounts: 26th March 2007
• 6 million registered accounts: 1st May 2007
• 7 million registered accounts: 5th June 2007
• 8 million registered accounts: 13th July 2007
• 9 million registered accounts: 24th August 2007
• 10 million registered accounts: 12th October 2007
• 11 million registered accounts: 20th November 2007
• 12 million registered accounts: 20th January 2008
• 13 million registered accounts: 27th March 2008
• 14 million registered accounts: 14th June 2008
• 15 million registered accounts: 4th August 2008
• 16 million registered accounts: 10th July 2009
• 17 million registered accounts: October 2009
• 18 million registered accounts: February 2010
• 19 million registered accounts: 23rd April 2010
• 20 million registered accounts: 27th July 2010 (1:55pm SLT)
• 25 million registered accounts: 28th August 2011
• 30 million registered accounts: 8th July 2012 (11:55am SLT)
• 36 million registered accounts: 20th June 2013
• 37 million registered accounts: 12th February 2014
• 38 million registered accounts: April 2014
• 40 million registered accounts: December 2014

Total Residents Today: 61 ,043, 239 - 7th October 2019
Daily Concurrency Today ranges from 43, 000 (max) - 26, 000 (min)

 

[Anya Ristow]

Tattere Nino in former times had a bunch of statistics - but it seems no archive.

Tyche Shepherd then again offers the daily concurrency count via API - so probably has an archive. But Tyche joined in 2007, so quite sure nothing before that time.

Tyche's numbers don't go back to the beginning. I used Tyche's numbers (the whole data set) to compare the *shape* (not amplitude) of concurrency plots from the beginning of when Tyche's numbers began (2007, iirc) to when I made the posting to SLU (2010-ish?), and I found that they plotted nearly line-on-line over season changes, school years, superbowls, daylight savings changes, etc, over the whole time range. Meaning, they took a 2007 (or earlier?) concurrency plot from a single week and scaled it for every week thereafter. That is, it wasn't real data. My theory was that there were so many bots that all the humans could log out and the concurrency graph wouldn't be affected much. I said they were likely duplicating/scaling a week in 2007 (or earlier) to simulate human activity. I also said that the only way bots could be run so consistently over so long a time was if LL was running most of them, signing them in/out to meet a target concurrency plot, and that they likely weren't actually bot clients communicating with the servers, but more likely just part of the server programming. You folks told me I was wrong. Numerically, though, the concurrency data 2007-2010-ish can't be real.

 

[Kamilah Hauptmann]

Yeeeeah.....noooo, not anymore.

You mean they've done some work since 2.49b?

 

[Khamon]

March 18, 2005 sealed our doom.

 

[Chalice Yao]

You folks told me I was wrong. Numerically, though, the concurrency data 2007-2010-ish can't be real.

2007-2009 was biiig. There's a good reason for it, too - 2007 was a year they allowed users to finally join without a credit card, just a few months before - which was a BIG boost to users. I joined in January 2007 due to exactly that, and those three years following were super active all over - folks exploring, sandboxes filled. It's probably the years when I met the most new people on average, so I'm not sure. Daniel's stats up there seem pretty accurate all around.

 

[Clara D.]

2007-2009 was biiig. There's a good reason for it, too - 2007 was a year they allowed users to finally join without a credit card, just a few months before - which was a BIG boost to users. I joined in January 2007 due to exactly that, and those three years following were super active all over - folks exploring, sandboxes filled. It's probably the years when I met the most new people on average, so I'm not sure. Daniel's stats up there seem pretty accurate all around.

October 2006 here. I was part of the hoards of free users that would ruin SL forever!

I paid the $10 for my first alt, even.

I think it spiked 2006-2007 because of media coverage, I learned about it from an NPR segment.

 

[Daniel Voyager]

I signed up to SL just after it reached 1 million residents with 11, 000 - 14, 000 logged in users (average back then).

Downtime Wednesdays and many grid crashes back then with viewer 1. No windlight or mesh back then either.

It reached 88, 000 which I believe was the highest peak ever for Second Life.

 

[Anya Ristow]

2007-2009 was biiig.

Not the point. The scale of the numbers isn't suspect. The invariability of the numbers is suspect.

For example, if you plotted January 2007 over top of July 2009, scaling them so they have the same peak amplitude, they were nearly line-on-line the same plot. You'd expect people to sign in later in the evening in the summer than in the winter, and they didn't. Or be online during the day more during summer than during the school year. You'd expect the *shape* of the plots to be different. And they weren't.

People tried to tell me Australia was in Summer in January, completely missing that (1) Australia is tiny compared to North America, and (2) Australia isn't even in the right time zone to balance North America, as the plots showed concurrency at time-of-day. I compared the plot before daylight savings, between North American and European daylight savings change, and after both changes, and they were nearly line-on-line. The expected result is the European peak should move an hour between those dates, and it didn't. Etc.

I never got anyone to understand this at the time, and I don't expect to get anyone to understand it now, without the plots But the bottom line is that I believe the concurrency numbers are made up.

ETA I didn't think of it at the time, but the charitable explanation is that LL never had the ability to report concurrency in real time and only pretended to be able to do that, and Tattere Nino and Tyche collected algorithmically generated data that they through was live.