The economy is false....

Ok, I haven't really slogged through this whole thread, but there's another thread concerning Chinese workers in SL here

Well, I'm glad to see someone finally getting around to post these story links.
I was getting SO upset as I read everyone accusing Greene Hornet of "racist"
comments. It was a few of these "Chinese sweatshop" stories that actually
brought me to Second Life in the first place! -- I'm intrigued in the notion
that soon people will wake up in the morning, and instead of driving to
work, they will hop on line and in to some virtual world, where they "do something"
to make in-world money... which is then converted to their national currency,
and THIS is the work they do for their employer .... 9-5.

I think everyone who made those accusations should appologise to Greene,
and Games Prototype, you're a big person to come out and do just that!


Besides the truth of the matter being that there ARE these sweatshops,
and some of you attacked him for making a true statement rather than
first asking for proof... there is a point I'd like to make about what makes
a "racist" comment. And I do NOT think that saying "There are German
officials and leaders killing off jews by the thousands" would be seen
as a "racist" comment in reference to what the Nazi party was doing during
World War II!

The fact is that some people do bad things... it's been that way since the
beginning of time. To point out that a group of people is doing something
you might not agree with doesn't make it a racist comment. He did not say
ALL Chinese are involved in running sweatshops! -- in which case I can
certainly see some people taking offense to that.
He said "There are [some] Chinese nationals..... "
Please people... get off your "politically correct" high horse where you make
it impossible to talk about what's going on. If there are stories out there
about Americans putting people to work in virtual world sweatshops, or Africans
or Russians, or whoever... why can't we just say that!? "There are American
companies that are putting people to work in sweatshops"... does THAT make
me a racist against Americans? -- or just stating a fact of truth?

Gabrielle

Actually I think Greene has a very interesting idea about prims and currency,
although he might not be expressing his idea very clearly... and maybe I
don't have a full grasp of it myself, but let me try to describe what I think
it means in a different way which might help...

Right now prim ownership is based on the amount of land you own, but
this does not have to be the case.

Right now it IS the case because a sim (computer server) runs a portion
of the virtual world and only has a certain amount of CPU power for all
the calculations needed to manipulate 15,000 prims.

Lindens charge us for CPU power which is directly tied to land right now,
but that is really silly. It means Second Life will not be able to have a
HUGE party with 2,000 residents and TONS of cool prims all close together
because they would all be in the same sim which can't handle all that CPU
requirement without some massive amounts of lag!

Bad, bad....

So, what if LL tied our monthly payments (or tier) directly to how many prims
our avatar can control?

Right now LL is charging $200/month for use of a sim (island).... which gives
you control of 15,000 prims.

WHY does this have to be tied to land ownership??

Why not just change some Second Life programming, so that if my account
is paying $200/mo then I can manipulate 15,000 prims, which an entire
server would be dedicated to. It doesn't have to be based on land. If I were
to want to control more prims, I could pay $400/mo and then be "renting"
two servers from LL... which would allow me to control 30,000 prims.

The software would be designed in such a way as to allow me to drop any
number of my prims in area of the world.... it doesn't matter what part of
the map I have prims in... the servers I am renting will be incharge of rendering
my prims.

For those of us who don't pay that much.. like the $10/mo lowest level premium
accounts would be allotted a certain number of CPU cycles for their prims.
Right now a $10/mo account can control 117 prims (if I remember correctly)...
and several of these lower level account would share the processing power
of a single server.... all paying a "portion" of the total rent LL charges for our
server usage.


So, back to Greene's idea of a prim-based economy....
As he said... it's not that prims would replace $L, but rather that the incentive
based programs like the rating system could be set up to allow control over
extra prims. Let's say I'm a $10/mo account... this allows me to control 117
prims. If I'm a good girl and people like me... rate me nice... and such, then
rather be awarded in $L, we can be rewarded in more prim control, so maybe
they'll let me control 150 prims for all my positive ratings.

My prim allottment might also mean I can rez 117 prims a month (or some
multiple like x2, x3, etc)... and so by having higher ratings I can rez more
prims each month.

The alternative would be to move to a higher tier, paying a higher monthly
fee.. which allows me to rez and control more prims each month.
And so "winning" or being awarded extra prims saves me cash, as I would
not have to pay more to get a higher prim usage for my ratings.



Now, Greene, even though I am ALL for the ideas above dealing with the renting
of prim control as opposed to tying prims with land... I'm not sure I'd like to
"get paid" in prims, because I'm so new and don't own land, and don't have a need
for prims constantly rezzed it does me little good. I AM learning to build, and
so I am rezzing a lot of prims. For this to really work though I would have to
be able to convert my prim allotment back to $USD.... which I am not sure I
understand... if you've even talked about a way to do that. Can you expand?

Gabrielle

Yes, Gabrielle - Yes!

You have captured (and expressed) my thoughts exactly!

Regarding the end of your post - I never intended "payments" in prims since I did not mean to propose that prims be hard currency. So the linden currency would still hold, and payments would be made in that sense.

But what would be important is that Linden peg its costs of operating a server (including economic profit/loss) in US$ to the Linden/US$ exchange rate. Since prims control our sl experience in-world then LL needs our real input on how to architect that experience in RL systems. The present dedicated land system is scaled horizontally - meaning a fixed number of prims per server. Why can't that number go higher? Can it be vertically scaled through multiple cpu boxes? How would LL know we are willing to pay for a big party with lots of people and lots of prims in the same place in-world if the price/prim in-world was not able to rise enough to purchase bigger boxes in RL?

And thats the key mechanism - stopping the growth in land (horizontal prims) and letting the price per prim float in-world while the linden/US$ rate floats in RL. Those two things would let the "invisible hand" work in SL and RL together for a better experience in-world for what we decide we want to do (collectively) and for what the Lindens really want in RL - a successful venture-funded startup. Everybody's happy and no one gripes in forums about what the Lindens did or did not do regarding land auctions, currency sinks, or tier fee changes. And in-world, the reputation system is transformed from "free-money" for vets to a +/- (good/bad) band (+15%/-15%?) around membership level/tier. And other incentives cascade down from there - including farming, etc.

Does this make more sense, expressed this way?

Why does the programming need to change? This is exactly how it is already. One sim rental ($195mo.) equals 15,000 prims, two sims ($390mo) equals 30,000 prims.

If you give bonus incentive prims to someone, someone else is going to have to go without. Possibly someone already paying for the sim quota of prims. Do you just say, "Sorry, your $200 only buys 14000 prims this month because your neighbor was a good girl. Maybe next month her ratings will go down and we can let you have the 15000 your paying for." This also means if I base my building ability on my chance for incentive, I am probably going to have to alter my build every month as my incentives rise and fall.

So, I still dont get it. If we have to pay for prims instead of land, how do you control the land use? What is to keep the prim rich player with 2048m plot from bullying the prim poor player that has claimed 16,000m of land and is keeping it bare, because they dont want to buy more prims? Does the prim guy have more rights to more land since he is paying for the prims?

Please dont say that if they pay for 117 prims they can have 512m land. That puts us right back with how it is now, paying for 512m and getting 117 prims. Big land would be desireable for the low prim user, simply for privacy, build a little house in the middle of a huge buffer. Why not? If I only had to pay for prims, I would want as much land as I could get my hands on.

How do you ration the land in a prim based tier? Each sim has a limited amount of land, so what is to keep a prim poor player from just buying up a whole sim worth of land? Will we have to pay for both prims and land? I apparently missed something very critical, help me out.

"Land" is just an abstraction - it actually stands for consumption of scarce resources (CPU power and bandwidth mostly, database space secondly). One of the scarce resources available are prims. So, tying "value" to an abstraction encompassing a scarce resource makes a lot of sense to me.

A prim-based economy would actually have two scarce resources - prims and the space to rez them. I'm unable to foresee what consequences this would have, beyond making everything more complex.

Actually, there are more scarce resources - like, say, scripts consuming CPU power (currently possibly more important than prims...). LL has tentatively thought about ways to "limit" script usage, or at least heard suggestions about residents proposing models of limiting script usage - again, relating to the abstraction "virtual land". So, bigger land would mean more available CPU for running scripts. I imagine that under your proposal, script CPU power would also be charged separately.

Another scarce resource is the number of avatars in a certain area. Again, I suppose that you should get charged for the number of avatars available on your land. After all, one avatar corresponds to some thousands of polygons that must be rendered at your graphics card, so, in a sense, if you have more avatars (and more polygons to be rendered), you should pay more for that privilege.

Particles are another scarce resource. Assuming you have your setting to "4096", this means that you should pay to have a bigger share of those 4096 possible particles being drawn at other resident's screens.

Also, if in the future one scarce resource changes, you would need to pay for it separately. I'm thinking about meshes replacing prims one day (not likely in the next half-decade or so ); LSL being replaced by Mono-based scripts; etc. So, if you place all your efforts on a prim-based economy, you must be prepared for the day it simply disappears from the face of the (virtual) Earth. As in RL, the most successful enterprises are the ones that don't resist change, but adapt quickly.

And so on. So, if I understand this idea correctly, all scarce resources, after a while, would be charged in some significant way. And the big question is if we should go that way, or just keep our convenient abstraction -- land.

In the ISP world, both approaches have been used successfully. Some ISPs charge for bandwidth, disk space, database space, number of mailboxes, etc., all separately, and you have a fine-grained detail in your billing, being able to pay for exactly what you need to have. Others prefer to offer an abstract concept, a "pack", which allows you a certain amount of "scarce resources". If you need more resources, you buy another pack, or increase your pack to the next level, or something similar.

Since both models exist in the Internet, I'm not sure which is "better". I'd say that as the number of Internet users still grows, there are more ISPs offering the "pack" system, because it's easier to understand for some customers. It's a simpler model, and, in terms of the company providing the service, they're better off explaining what the "pack" has, and deal easier with it.

Now in SL, LL has opted for the "pack" solution - virtual land is tied to a certain amount of prims, a certain number of avatars that can roam freely around the sim, and, eventually in the future, a certain number of scripts you may have running in your land, and so on. So, the whole burden of assigning "value" to the separate scarce resources is "transferred" towards the abstraction, "land". I agree that the biggest disadvantage is the high level of discounts in tier which makes prims much cheaper for big landowners - but if you follow the Town Hall meetings, you'll see that LL is definitely rethinking the way tier discounts are applied, and have recognized that there is undue disadvantage for residents owning large plots. They also haven't been totally consistent in their "abstraction" - thus, you pay separately for uploading textures or animations (mostly because when you do so, you're not necessarily doing it at your own plot, but cn upload them anywhere in the grid), and will in the future pay separately for using XML-RPC, if I understood that idea correctly.

I think it's only a question of personal choice. When describing SL to others who never joined this virtual world, it's far easier to explain the concept of "virtual land" (although even so it's often misunderstood, as a somewhat recent thread in Slashdot has shown) by simply pointing out that it's an abstract representation on the resources you're able to use in-world - bigger plots meaning more resources available. Explaining what exactly a prim is, and how much it costs (and why!) is far more difficult, to someone who never worked in a 3D environment before, and has absolutely no clue how many faces a cylinder has, or why a torus creates more lag than an untwisted cube.

At the end of the day, the point is, if we prefer to deal with a single abstraction -- virtual land -- or with dozens of separate things that you are charged for. I can only say that for a non-technical person, without any clue on how the servers work, the "virtual land" abstraction is easier to grasp, and allows for a better identification with RL, where land is also a limited resource which allows you to build upon it and fill with items...

*where* would one rez their prims? When you rez a prim, you put it in a specific x,y,z location. Are you saying all of second life would be a giant sand box?

And you think people would like that better?

Besides, you can already rez all the prims you want in the sandboxes. Why would anyone buy a premium membership when you can rez all the prims you want anyway?

I don't think you are thinking it all the way through.

Buster

*Applause*

Bravo

I'd just like everyone to know that I am currently in a roomful of 200 albanians who are chained to their computer desks wandering about SL harvesting money trees.

And yes, I am holding a whip.

(apologies to anyone from albania)

Loki, thanks for such a long/interested reply, but yes... I do think you are
still missing a vital point. You are pointing out that our $195 allows us to
control 15,000 prims (along with an entire island/sim of land).

What if I don't care about owning land at all? The reason MOST people
buy land right now is not actually to own that much space they can build
something on, but rather they buy land because it allows them to control
15,000 prims.

So, Greene's idea (an mine).. is that LL should not tie $$ to land, but
rather tie $$ to the amount of prims each player has rezzed.

Having a sim worth of vacant, unused, un trafficed land does not take
much CPU or resource power at all. A single CPU could probably handle
quite a bit of "land" if no items and no people were there. What uses
CPU/resources is avatars, prims, scripts, etc.

So, the idea is that someone pays for the resources THEY use.
Not the land owner. The other day I was told by a club owner to
take off some pretty particle item I was wearing, because it was
causing undue lag. WHY should I have to do that?? I bought it
to WEAR it! not hide it in my incentory.

The problem is that right now the land owner is limited to the CPU
resources they control (15,000 prim and other resource limiting stuff
like scripts, etc).

The solution is to have each player responsible for their own items.
It works by having a different CPU (not the one which the land is on)
responsible for handling my avatar, scripts, attachments, etc.
If I wan't to have all kinds of attachments that require CPU power, then
I am told it will cost me $L2/hr to wear this flashy attachment.
If I want to rez a super fancy house, I'm told it will cost $L5/hr to keep
this house rezzed.

The land owner no longer has to care, because I'm not taking away
from any of THEIR resources. My flashy clothes use a different CPU
to handle my avatar and all rezzed items I own.

For land owners this is GREAT! Because now they can have 1,000
people in their club at a time because each avatar is controlled by
some other CPU dedicated to that process. Residents love it because
they can be involved in bigger parties with out lag. They also love it
because they don't have to give up their fancy/flashy CPU intensive
items... as long as they're willing to pay for the privilage.






Nope... your $200 gets you access to control 15,000 prims. If I'm a good girl
and get to control more then my "normal" share, then a different CPU can be
responsible for keeping my items rezzed. Stop trying to tie actions/items on
a specific piece of land to a single computer... THAT is the key.

If one computer can handle rezzing 15,000 prims, and you're paying $200/mo
then you get an entire CPU to yourself... which will rez and handle your prims
no matter what parts of the world you or your prims are in. Your CPU controls
your stuff... not the land owners.

If I'm paying $100/mo and two other people are paying $50/mo, then then
three of us can share a single CPU which is responsible for rezzing/handling
out items... 7,500 for me, and 3,750 for each of the other two, for a total
of 15,000 which the CPU can handle.

Doesn't matter where we rez our items... doesn't matter how much "bling"
or particles we use... as long as it doesn't go above our alloted limit.
The land owner sees NO lag because their CPU is not responsible for
processing our items... our CPU is.


Gabrielle

People would like it MUCH better if they could join in larger
parties without lag, yes. People would like it much better if they
did not have to take off bling or particle items because the CPU
responsible for the landowners event could not handle the load.
People (landowners) would like it much more because they
could now have tons of people at there events.

As long as LL has CPU resources tied directly to a specific piece
of land, this wil be a problem.

If I pay $200/mo that allows me to rez 15,000 prims right now,
and would also with the new idea.

The difference is that under the current programming, the 15,000
prims I can control are the prims on a specific piece of land...
and there is no way to allow a piece of land to handle more than
15,000 prims, because each server is responsible for the actions
going on in that piece of land.

The new idea says that if I pay $200/mo and can manage 15,000
prims, it does not matter where in the world those prims are.
It's not tied to any specific land. I can rez items in any place that
allows me to rez items, but it's MY CPU ($200/mo worth) that
handles the processing/rezzing of my prims... not the land owner.


If someone rezzed 10,000 prims in a sandbox right now, that
effects all the people in that whole area... they see lag, and if
it reaches 15,000 you've limited others from building there too.

The new idea says if I pay $10/mo I can have 117 prims
rezzed at any given time. A small portion of some other
server is responsible for managing my prims (along with others).
This way I consume only the CPU resources I'm paying for.

Linden Labs might make more money from this as well...
since, as you stated.. you can rez all you want in sandboxes..
so why should I ever pay more than my first $10?

Right now I can get in-world, rez a house 50 meters above
the sandbox, rez all the items, vehicles, toys I want....
consume all kinds of resources and never pay anything more.

However, if, as a Basic Account, I was limited to 50 or 100
prims at any given time, then I might be more motivated
to pay $10/mo for an extra 117 prims or so.

Those people who just play tringo, chat, etc and don't
consume a lot of resources will still have their option of
a pay-once account for life.

Gabrielle

Gabrielle, I wonder how technically you can get "another" CPU to handle per-user tracking of prims.

I like any idea that makes people pay for the resources they use - that makes all the sense to me. There was brief talk once about LL charging us (either in L$ or US$) for the prims we attach to our bodies. Lately, however, LL seems to prefer to do the rendering of attachments more efficient and rely upon Havok 2 to process things faster, so I think this idea was abandoned.

Even if technically you could have "CPUs allocated to residents" - I'm still struggling with the concept - there would have to be some way to overly control the interactions between objects. Two avatars meet in a club and each touch a dance machine. Which CPU is going to handle the scripts inside and the loading of animations? Both? The CPU "tied" to the club owner? These things are tricky, although some platforms - notably, OpenCroquet - tend to favour this idea. Basically, it's your own computer's CPU that handles your own objects. Naturally, this would change our experience of SL greatly. And in that case, "charging" people for using their own CPUs would not make much sense.

Since I wasn't around on the days of the prim-based economy, I can only understand it was abandoned for some good reason. Like charging for using telehubs (which are now for free).

As said, I'm definitely not against the overall idea of charging people for the resources they use. I always defended, for instance, that scripts should be "charged" as well as prims - these days, they're able to create far more lag easily than prims. But I also understand that "measuring laggyness" is something too abstract and very hard to "charge". A one-line script can lag far much that 50-100 very long scripts which are cleverly done. Which should be charged more?

I leave with more questions than answers...

Just to inject a little reality into this discussion: There is no way that SL is going to be rearchitected so that you "attach a prim to a processor". A possible future involves the sizes of sims changing to reflect the complexity of their contents. A likely future involves faster or idle servers automatically switching in to support high priority or highly-loaded sims. From various comments I believe the current state of the world is that sims can be manually moved between servers, with a small service interruption.

And honestly, all SL needs is the "likely future". Add a little big iron to the grid, improve the interconnect speeds, shuffle servers between sims as needed, and we would be fine. (The asset server(s) should already be pushing mainframe status with dedicated interconnects, as distributing them safely is challenging)

Simply applying prim count or script cycle limits to accounts is independent of any changes in server architecture. Both are fairly easy and I would not really be surprised if one or both happened in the future.

The most relevent question at the moment is: Do we have land that is unused soley because someone bought it to put the prims elsewhere? Then it would make sense to allow them to increase their prim limit without buying more land, it all pays for better servers one way or another.

I'm enjoying this part of the discussion!

What if there was no database? What if everything resided in in memory? Why do we need a transactional architecture anyway? Couldn't it be a platform based on telco services, with a high-speed messaging bus/broker to enable in-world interactions between virtual machines in memory?

Big Iron is classic transaction processing with a database back-end - works well for an online store but not a virtual world that needs real-time, one-to-many, many-to-one, and many-to-many interactions; that scales vertically using multiple processors, and is also fault-tolerant, redundant, etc. like the old Non-Stop or other high-end TP stacks. OS is part of the equation, number of processors, memory, messaging, and database are the others. All of this can be abstracted by hardware these days, right down to the physical network elements a la today's major carrier networks running IP.

The "land" abstraction is convenient - for quick, summary explanations, but it does not work well financially (under-ulitized physical assets resulting in low ROA - a killer statistic in any service business model), produces lower quality of service (just imagine how tolerant you would be of a telephone call that lagged, or how frustrating it was when the "circuits full" message used to play when making a long-distance call), and has no direct, economic link to the primary scarce resource - prims.

The limitations for scripts, messaging, etc. are all supporting features to prims since the prevailing metaphor defining our in-world orientation is physical "land". In real-life the differences in land values hinge on distributive effects - what's in it, what's on it, where it is, what you can do with it, how big it is, who owns it, etc. Improvements to real land increase its capacity to yield, or produce greater "rents". In SL every piece of land has a fixed number of prims allotted - 0.2285 prims per sq. meter - that number never goes up or down, no matter how much you build, no matter how well or how badly you behave, no matter how much you make in Lindens, or how much you contribute to the greater SL community. Some would say that's "fair", and yes it is good horizontal equity (treats everything similar the same). But it hurts us too.

By fixing the number of prims per land the Lindens grow our world by discounting prims to large purchasers - who most would agree are speculative buyers at this point (some might say Linden "partners". We don't pay the same price in real currency for the same number of prims, and the expansion of the world decreases overall operating margins for LL as revenue grows at a decreasing rate. I would wager a guess that the large land purchasers and island estate owners are the principal focus audience for every new SL improvement, despite the fact that the other 98% of the population in-world produces higher margins per prim. A fixed number of prims per land makes SL an upside-down pyramid with regard to quality of service and is not scalable financially.

The other problem is in-world - why can't every resident attend an announced town meeting? Is it still a "town" meeting, or just a token meeting? What if everyone wanted to go to the same casino? Why can't we? Why purchase computing power if I can't move it around SL? Could we all attend the same meeting in a Sandbox? Why should scripts slow things down?

These are vertical scaling events. They depend on scalable computing environments that move in tandem with requirements vertically. Since these cannot be pinpointed in advance they must be implemented through abstraction - not a one-to-one correspondence between box and place in-world. Think boat-loads of RAM and multiple CPUs in the same box, think software virtual machines, think highly distributed, think real time, think fault-tolerant, think redundant, think vertical equity in-world (treating different things differently).

The concepts in this case are less important than the practicalities. I'm simply asking why can't prims be bought and sold at a fair market price, the same way land is traded today? If the price of prims was allowed to float then the Lindens would get a better idea of whether they should expand horizontally or vertically as they grow our world. If everyone paid the same going rate for prims in the auction and secondary markets then there would be no inequities between residents. By uncoupling the pricing of land and prims then the "invisible hand" could determine whether large plots with no prims or small plots with large amounts of prims were desired. Similarly, in RL I can sell the hunting, mineral, or air rights to my land at the going market rate. These sales reduce my potential rents from the land but if I only intended to raise my family on the spot then I've gotten all the use I personally wanted. Why can't I buy a fancy piece of furniture bundled with the corresponding prim capacity if I want to? On the other hand, how do the Lindens know my real demand for prims (aka future revenue stream) when I buy land that has a fixed ratio of prims? Is that any way to run a business? Would LL create SL if RL worked the same way?

I know a lonely housewife in Florida who works online in SL to accumulate Lindens in order to sell them for US Dollars. She'll do anything to get the money. So, is the issue that the Chinese and my Floridian friend should be allowed to do that but for prims or is it a moral issue that these so-called Chinese sweatshops should be punished for exploiting these kids by screwing everyone else over and changing our currency? I'm confused because your post is all over the place.

BTW -- what happens when the kids turn 14? Do they move to the Nike factory?

Prims are as important to me as the money is. so prim rewards will not work for people like me. I am not alone in this many more people just make stuff in the sand boxes and sell them wile keeping land tires low a 512 spot of land is just fin for them. I don't care about prims.

I admit I did not want to read all to posts here and just skimmed through them.

oh And Idon't care about the sweat shops either.

As well as financial ones - which is the main point. Every economic system has distributive effects which may be separated out into horizontal (equals treated equally) and vertical (differences treated differently) equity.

My main point is that the present practice of divvying up prims at a constant ratio to land (sq meters) is fair in one sense (treats everyone the same) but unfair in another important way (lets some pay far less for prims than others). This second, more important aspect is vital to LL financial success as well as our in-world experience (quality of service).

The present incentive system focuses on a currency and an exchange ratio that is managed, or in many respects artificial since it is not tied directly to computing capacity, or prims. The ability to trade prims separate from "land"; that is, to price prim capacity in Lindens and to make them portable across property lines or sims would align the incentive system with the scarcest resource (computing capacity), let residents signal LL on how SL should be developed through behavior, and make the Linden currency a "hard" currency less likely to be farmed - in FL, China or anyplace.

Linden currency farming, regardless of the moral implications, is a financial drain on LL (excess computing capacity), on residents (removes currency from circulation in-world), and exaggerates/distorts in-world economic statistics which may be used to determine future SL development efforts or services from LL and others. As a metaphor you can think of it as a giant siphon on our air supply into space.

Is this more straightforward?

The SL economy - since the L$ is freely bought and sold on GOM and IGE , further more Land being teired / leased in $USD ....

is directly related to the Real World economy.

Demand for content is an entertainment expense.

Supply of content is an income for largely recreational pursuits by those skilled in it. It is of course time consuming and potentially challenging, but the work is mainly for fun.

many have argued to use the USD$ - i dont agree becuase of the Tax implications to transactions that never leave the SL environment. But realistically its money changing hands.

Paying for prims is an artifical construct - prims are only one part of content anyway.

I guess my point is the SL economy isnt truly a seperate economy at all - it is a platform where people buy and sell services - LL's responsibility is to try to keep the platform stable, thats it.

And so if you've read Collette's post above this you will see that we've gone full circle...

Its now time to close this thread and move on (I wish I knew how).

Well, I'm not a coder at LL, so I don't know exactly how
they are implementing things right now. From the little I
do know is that it seems things are set up in such a way
that ANYTHING that happens in a sim is handled by that
sim... although I hear talk about an asset server which
might handle inventory... but that's kinda vague.

Anyway... I assume that while Joe Avatar and Jane Avatar
are in the Morris Sim, then it's that one CPU responsible for
crunching the numbers, collisions, line of sight, displays, etc
for each resident... Get enough residents in a single sim,
and the CPU can't handle the load.... everyone there gets lag.

The new idea is that each person is alloted a small amount
of cpu usage (which we're all getting right now anyway)...
Why does it have to be the server which displays the land
we are in that crunches the numbers for our avatar? Why
can't the actions of Joe and Jane be (potentially) managed
by different servers, that interact with the Morris Sim to get
some values, but then the seperate servers crunch the numbers?
This would free the Morris Sim itself up from lag.

All free ($10/life) accounts are given a certain percentage of a
server/cpu they can consume. If they try to rez more items,
view with more detailed distance, attach 5 million items to their
AV, or anything that tries to take up more resources, then
they are not allowed... and something has to give.

OR.... they could pay an extra $5/mo for additional CPU
power which could go towards rezzing/managing more prims,
or activating more scripts, etc.

The bottom line is that this does not have to be tied
to land ownership... really it's all about CPU resources!




Well, if LL could find a way to put more on the client that
MIGHT be an okay thing... certainly for them it would be
good because they could get away with needing as many
servers to crunch the numbers.... for residents it might be
bad because people might need to buy even BIGGER
computers for home... and that might limit the number
of people who can join SL.

... I'm not suggesting this.

I'm suggesting LL servers still crunch the numbers, but that
different/new servers would be available for rent ($5/mo for
some amount of CPU power) that each person could use.
It would be handled in the same way as land is handled now,
but we would not HAVE to tie land square meters to CPU power.

As far as what if two AVs meet in a club and click a dance box.
I would say this could be done either way....

1) the club owner's CPU being responsible for handling the scripts.
With the mentality being that if they want to privide dancing
for their patrons, then they should bear the burden, pay for
the extra cpu needed, etc

2) But I think I prefer the 2nd option... where it's not the club
owner, but the individual. If an individual wants to have the
full experience of SL, then they should pay for the level of
experience they want to have. If they want to wear 1,000
prim attachments with bling and rez 5,000 objects on 512m2
of land... then let them, but they will have to pay the extra
money needed for the amount of CPU power to handle all
that.

Right now SL is set up that you rent CPU usage as "land"
and connected land in a sim can be joined to allow more
prim use.

The future should be that SL is set up to rent CPU usage
directly.. without the need for adjoining land in a single sim.

If I want 1,000 attachments... let me rent that much CPU
and let it be managed by the DIFFERENT server responsible
for my AV... not the land owners.


Gabrielle

Though your idea is nice, it's not reasonable.

What you are suggesting is simply not feasible. Two server CPU's can't cooperate that way, it takes an amazing level of intricate optimization to make it work within one process. Interprocess communications during graphics rendering? You may as well call up Ford and suggest they simply double the gas mileage of all their cars. Great idea! Why didn't I think of that!

Many of the calculations you are talking about are on the client side anyway -- CPU time is only one of the factors. The CPU runs the LSL scripts, brokers the assets, calculates physics, etc.

The overhead required to administer CPU time allocation like you suggest isn't feasible either. It is quite easy to imagine, and quite impossible to program within the constraints of making everything else work.

It is an age-old problem on time sharing systems that the computer spends more time managing the allocation of the CPU than it spends calculating. The scheme you suggest could easily use up twice as much CPU time just keeping track of how to allocate CPU time. You would have the technical capacity to administer LESS prims per unit of hardware. The current scheme solves the technical problem of how to allocate CPU time without wasting any CPU time doing it.

In fact, LL is actually going in the opposite direction from what you suggest. There are plans at INCREASE the affinity CPU time to land ownership, so that larger plots get more CPU time for LSL scripts. Tying server resources to land size solves a myriad of technical problems and eliminates the overhead of doing it any other way.

Buster

P.S. Sorry to disagree with you on the same issue in two different threads on the same day. That doesn't mean I don't love you.

Greene,

First off, I confess that I haven't read most of this thread. I find your conclusions about SL's vertical scalability very interesting though. I just pulled up the map and turned on telehub coverage.

I don't have the energy to evaluate the whole grid, but JUST in the sims serviced by Waterhead telehub I can count 17 sims with NOBODY IN THEM AT ALL! Additionally, the sims in this area of the map that do contain residents are nowhere NEAR capacity! What a phenominal WASTE of (RL) resources this must add up to grid-wide!

How long can SL continue to operate under this archetecture at all? It seems to me that a more integrated solution like a single, highly scalable SGI or Sun server could handle the current (right now) load in the sims serviced by the Waterhead telehub using far less than 1/17 of the computing resources currently demanded by them! I think that this should frankly be ALARMING to anyone paying for access to SL.

As an additional aside, if a huge event were to happen in one of these sims right now with an integrated solution offering the SAME computing resourses, it is likely that 30 times as many people could attend. This is from the perspective of server resources though, who knows how the client would cope with that many av's at any rate.

I could be way off-base as I don't know anything about how SL is developed beyond it's general archeticture of putting only 1 or a couple of sims on very small servers. This is just my observation/opinion.

Major problem with moving to big iron is cost: taking a quick peek at Sun (one of the few with easy online prices), a dual-processor blade is ~10K, while a single 24-processor server will set you back more than 750K.

My bet would be that they are going to try and beat down the problems at sim borders. These are non-trivial, but far from impossible to resolve with a static grid. Once the sim borders are not a source of problems, a mid-term solution would be to split highly-loaded sims onto multiple machines.

I suspect overwork is more a factor at the moment than incompetence or lack of foresight. LL is definitely not a very large company, and they are working to scale a non-trivial system that I suspect never quite worked correctly.

<edit>Guarantee that the sims that are most often idle are often are the ones that are combined, so it is not quite as bad as it looks. AFAIK moving sims at the moment causes a short loss of service. Once sim borders are not sporatically lethal and sims can be rearranged invisibly, LL will have a lot more options. I'd call them idiots if they decided dynamic or predictive adaptation of the grid not coming, but it is not here yet...</edit>

<edit id="2">Yeah, me too with the edits. My point was that even buying from Sun (which gouges rediculously for blades) the cluster is 120K while the single server is 750K. Assuming the 24-processor server could handle the entire grid, a cluster a third the price would almost certainly be better. The cluster drawback being that it makes the software harder to write...</edit>

My concern is for all of the resources sitting totally UNUSED, wasting away, not the rare overuse. It just seems to me that they were using about 17 times as much compute power as they needed to be, at the perticular region I looked at, at the perticular moment in time that I looked. When they have a log-a-thon 5000 event that can only generate under 4000 concurrent connections, I just don't see the point of 500 or 1000 servers. It seems to me like a failure of the archetcture.

edit:
I guess I'm just saying, wouldn't it be great if all that unused capacity could be rerouted to where it's actually needed? Then we'd need a lot less CPUs on the grid overall.

edit 2:
It's not beyond the realm of possibility that the single 24 (Sun) CPU solution u mention would run the entire SL grid.

edit 3:
Sorry about all the edits, LOL. I also wasn't trying to suggest that anyone was incompetent. SL is a technological marvel. It just seems like it's current setup is a bit wasteful of hardware resources.