CTO Cory Linden Leaving Linden Labs

Specs are off a little there..

From http://blog.secondlife.com/2006/10/16/looking-forward-to-class-5/

The sim code is designed to run one sim per server core - the twin dual-core servers Ian Linden talks about above can each host 4 sims and earlier class servers, having fewer cores, could host fewer sims. These are not state-of-the-art processors and don't have tons of memory - any recent home machine should, in theory, be able to run one.

Of course, the above totally ignores bandwidth requirements, especially the upstream part..

This sounds like "void" aka "Open Space" sims. They are available now, in sets of four.

This whole "roll your own sim" discussion is rife with magical thinking, though. Why would LL allow interconnect to their grid without a charge nearly as high as that of leasing a sim from them? As we all know, hosting is a commodity, and LL's is too--that's *not* where most of tier or estate fees goes now, and it won't be where the interconnect fees will go in the future. Bringing your own machine to the party is very nearly irrelevant to LL's costs, so it would be a strange and stupid business plan for them to deeply discount grid interconnect fees.

One may *wish* that hosting one's own sim will save lots of money, but that doesn't--and won't--make it so.

There's also currently no evidence to suggest that thrid party hosting services would be any less expensive. They may be less, but I think it's doubtful that it would be by much. If/when we see them, they won't be charities.

They may, however, be competing, which could have the effect of lowering prices.

Personally.. I would hope that if the client and server are open sourced, then somewhere a company or an individual will produce a version which removes the geographic basis of hosting entirely. There is after all, no absolutely unbreakable reason why the area a single computer hosts can only be a 512x512 square that's all contiguous - it could be 1024x1024 with the same number of prims, or 4 256x256 squares scattered around the world, or similar.

The concept of a distributed network of sims, which are no longer owned by LL, is something that has been in discussion by both LL developers and the open source development group.

I don't know the current status op 'opensim' (opensimulator.org), but generally, the open source community didn't wait till LL 'specced' the protocols and APIs necessary to implement this, but started developing their own servers, based on current protocols available. Although LL developers and open source community have discussed this several times.

The biggest problem with a distributed network of open simulators, not owned by LL, is inventory and assets. Not just from a technological standpoint. 'Opening up SL' and making assets shared over the internet will have major effects on security & permissions and in that way on the economic/social model within SL.

To tie it in with the general question, why is Cory Ondrejka asked to go? I think this is one of those areas where 'technical solution/freedom of development' and opening up second life immediately clashes with LL's business model (hosting land). A key area which can cause a major strategic difference of opinion between CTO and CEO

Although i doubt this specific case has been playing a big role in the conflict (nobody except Philip and Cory knows the whole story, and because of legal repercussion, neither will probably open up in the near future). It's a nice illustration of mission & business model clashing.

I did mention skinning and buttons, the ball is already rolling on it, LL are coming up with a new UI, it just needs more of a push. However some of the changes that devs would like to make to the UI require changes to the server code, development is hampered by this.

It could of course be argued that development therefore needs to fit the needs of the current code but I think UI changes are important to potential customers.

quite off. haha. ok i apparently increased performance by 40%. not bad huh?

i did build a slick AMD machine for about $500, but it's al off the shelf. Custom server design exponentially increases the price. and yeah, that bandwidth. maybe we can all move to Korea.

PS: average business rates in SF for 175 watts of electricity are $0.04/hr ($28/mo baseline)

The real concern, for at least me and RL companies, will be stability.. Who cares if the thing costs US$20/month less if it's down a lot or is in some backwoods part of the internet with crappy pings to everywhere?



Try sending Cory a bill for the performance improvement - he's a short timer and might let it slip through..

bottom line: It's inevitable that bandwidth and storage size is going to jump astronomically and it's inevitable that it'll have to work as I saying.

Step back a bit and see what a joke it is to be paying 200--300 a sim. It's ridiculous. Imagine having to pay $300 a month to use Windows, OS X or Linux, or to get on the internet.

In the not too distant future one will look back and see how absurd it was.

back in the last century it did cost about $300 to have a decent web site until things got cheap. I actually consider SL like the web in the early 90s, but with a Z-axis.

No doubt Court, I'd say SL is struggling to exit what would be the equivalent to the brochureware era and entering the web app era. (He says with abundant amounts of hope in his heart)

What I'd be hoping to achieve by hosting myself is the ability to throw more and more resources at the problem when the need/ability/finances allowed/necessitated. I would be willing to pay just as much as the Lindens charge for a sim if I could host my one sim on my 4 quad core processor, 32 gig ram, zfs filesystem, in my third of a rack at my hosting facility, and get the comparable increase in number of AVs and prims per sim.

Sims, in my mind, provide plenty of space, it's the number of AVs and prims they support that is truely the gap.

As far as cost for the official LL servers opposed to the roll your own, even sending them your own box for them to host is not something they will be interested in. The achieve economies of scale through their purchase power and an increased efficiency through the standardization of hardware. Supporting 10,000 of the same exact box is much easier than supporting 1000 boxes if half of them are any box a client throws together and sends over.

No need to send any boxes.

Seems like that plus a "browser war" type. hence IBM partnership. If a virtual world wants to survive for more than a few years, it has to be open, especially if its running on open sourced technology. If standards are developed then embraced, and one virtual world doesnt follow them, it wont be a very popular virtual world.

I wouldn't be surprised if a new development team has spawned to create this. It would be nice to know more, as I'd love to contribute more to sldev, but i dont want to waste time on something that might be dead in the water. I still have Netscape 4 nightmares.

Indeed, as we did with everything. I used to use pay as you go dialup, now that was expensive.

However it is not inevitable that it will work as you're saying, there is talk of tying usage to the avatar, the more inventory you have, the more you pay is another solution, pay as you grow, if you will.

To quote Phil himself in an interview whose link is on the homepage

Q: If you do open-source, how will you make money? I’m sure there are a lot of developers waiting in the wings for this to happen, but can it be a win-win situation? Is an option like finding a middle ground between open API’s on one end and completely open-sourcing the code on the other a more realistic probability?

A: Because of the network effects that will tend to cause most people to prefer spending their time and doing their business in one virtual world, we think we can have a great business, even in a very open source/standards environment.

Just to comment on a few items:

OpenSim - you can dip a toe in the water today, by running the client
against your own server: http://wiki.magrathean.ca/index.php?title=DGiG

Storage/Bandwidth/etc - short term problem. Upstream speed is the
killer. Am reminded of the web in 1994. Virtual World hosting companies
on the horizon? Sure!

Future? - Eloise Pasteur has a nice bit:
http://www.massively.com/2007/12/14/the-future-for-virtual-worlds/

Management - There are Three Keys (CTO/VP Community/CEO). To me
it seems extremely important as to how they are perceived by the outside
world. It doesn't matter what I personally think of how they are doing their
jobs (in over their heads). What does matter is how 3rd party companies will
vote with their time, effort, and money, once they have an increasing amount
of Virtual World choice.

Actually. many home computers ARE fast enough to be servers. The current Class 5 servers have two dual-core Xeon CPUs and 4GB of RAM, and run four sims (one per CPU core). A typical current home computer (dual-core Core 2 or Athlon64X2) and 2GB of RAM should have no trouble at all running one sim; even running two sims, it would have about the same amount of processing power as LL's servers do. If you upgraded it to 4GB and gave the server a dedicated hard drive, you should even be able to do other things with the computer at the same time as running a single sim.

Bandwidth is another story. The average bandwidth use of the SL viewer is at least 100Kbps. (It varies wildly; it can be as low as 5Kbps when the in a quiet area where the user already has all the textures, or 1000Kbps or more in a busy area.) The typical upload speeds of home connections (384K or 768Kbps) won't go very far.

Quite so. The initial target market for self-hosted sim interconnect will be those (mostly corporations) willing and able to spend much more than the current cost of having LL host their sims for them, because of technical advantages they gain by running their sims on their own ultra-high-end hardware or with very special access to their corporate networks.

I agree that in perhaps a decade there may be enough of a market that it'll be possible to interconnect to any of the competing, interoperating grids, at which point the associated fees would be much lower, per connected server. But initially, there's just no reason for LL to price interconnect attractively to the "hobbyist" sim-owner market: there's no margin there.

Phil talks to the BBC
http://news.bbc.co.uk/1/hi/technology/7144511.stm

Aren't the captions on Philip's picture with his avatar the wrong way round...?

Broccoli

Suprisingly not.. His avatar is very well done and he looks a little goofy in RL..

Cory avie is a women right? but he a guy? Well that sums it up. Philip Linden found a new lover Thats why he was fired hehehehehe

Or Phil is actually a program in Second Life that has achieved AI status.. and uses an RL avatar?

And what does he mean that their differences was more about how to run the company? Like, the love machine and the Tao of Linden? Good grief, both of those suck for a corporate environment, especially in the areas of customer support and communications! But I thought Phil was really, really, really into those things. I mean, he's touted them in the blog often enough, even though we've pointed out the fallacies about them and ridiculed them to no end.

*shrugs* Either way. Good luck, Cory!