OpenSim is the metaverse 1.0

Someday soon, we will have a virtual reality metaverse. A fantastical place, limited only by our imaginations, where time and space mean nothing and experiences are shared as easily as bit and bytes are today.

What technology will form the basis of this metaverse? Will it be created by Facebook? By Google? By some startup we haven’t even heard of yet?

Nobody knows.

But we already have a mini-metaverse right here with the hypergrid. A metaverse 1.0, if you will.

Here’s why.

1. It is open source

The World Wide Web has become as successful as it is partly because it’s based on open source software and standards.

Yes, there are proprietary platforms that make up parts of it — Windows Server, for example, is used to run many websites. Apple’s Safari is the browser of choice for a certain segment of the population.

But the bulk of the World Wide Web isn’t owned by anybody, which helps make it open to everyone.

The hypergrid is based on open source software — OpenSimulator — and on the LibOpenMetaverse library. Its main viewers are also all open source.

2. Individual sites have full ownership and control over their content

Having your own OpenSim grid is a lot like having a website, as opposed to, say, a page on Facebook.

A grid owner has full control over all the content, users, and technology on their own grid. They decide who can visit the grid, and what other grids it links to.

Other than law enforcement, nobody else can dictate what happens on a grid. Just like nobody else can tell a person what can and can’t be on their website.

3. It’s infinitely scalable

The World Wide Web can expand indefinitely as new websites get added. And the fact that new websites are added doesn’t affect other websites — it doesn’t make them slower, or more difficult to get to.

The same goes for the hypergrid. A million new grids could be added tomorrow, and it would have zero impact on the existing grids. Although there might be an indirect impact, if these new grids also brought in new users, some of whom decided to visit the older grids.

The hypergrid. The large blue sphere in the middle is OSgrid -- unfortunately, still down.

The hypergrid is a universe of worlds connected by hypergate links.

4. With some basic technical skills, anyone can put up a site

The early Web was dominated by tech-savvy individuals, educators, and creative types putting up their own little websites. Over time, big players got into the game as well, but the vast majority of websites are still tiny niche ones, and the numbers continue to grow.

All it takes to put up a website — or an OpenSim grid — is a computer and an Internet connection.

5. There are many hosting vendors

Of course, grids — like websites — can quickly outgrow the capacity of a home connection.

Professional hosting is available for both websites, and for grids. A list of OpenSim hosting vendors is here.

And anyone can get into the OpenSim hosting business because the underlying software is free and open source.

6. A grid can come in any size

Just as you can have a single-page website and a giant site with millions of pages, so a grid can be any size as well.

There are single region grids, and grids with thousands of regions.

A grid could be a single little shop on an island all by itself. Or it can be a giant corporate campus, sprawling social community, or shopping mega-mall.

7. Third-party services are welcome

One of the ways that the World Wide Web was different from a single platform like America Online or Compuserve was that on the Web, different services were provided by different parties.

For example, in AOL, users got everything they wanted in one place — search, news, shopping, friends, personals, messages, and so on.

On the Web, they might get their search from Google, their shopping from Amazon, their classifieds on Craigslist.

Similarly, on the hypergrid, users can go to Sanctuary or Hyperica to find new destinations, the Kitely Market for shopping, Virwox’ OMC for currency. As the hypergrid expands, more and more niche grids will spring up to serve particular needs, and users will be able to pick and choose to find what works best for them.

As a result, no one organization will have to provide everything to everyone, and the speed of development of services will be high due to the competition.

Items on the Kitely Market.

Items for sale on the Kitely Market.

8. The technology lags behind

Yes, I meant to include this one. Having the latest and greatest technology platform is not necessarily a good thing.

The early World Wide Web looked lousy compared to AOL.

For years, sites that wanted to look really spiffy had to use proprietary add-ons, such as Flash, because the basic HTML was pretty minimalistic.

The benefit of this, however, is that the Web runs on a wide variety of devices, including some really old computers.

Similarly, OpenSim doesn’t have the fanciest graphics. But, as a benefit, it can run on older machines.

9. It allows for user-generated content

The Web didn’t just grow from people putting up Websites. It also grew by people uploading content — vast quanties of content — to existing sites. Puppy pictures on Facebook. Introspective posts on blogging sites. Status updates on Twitter. Restaurant reviews, auction listings — the list can go on and on.

In OpenSim, users can also create their own content. They can upload photos, and build stuff out of prims right inside the world.

The Second Life-OpenSim ecosystem is one of the few platforms that allows in-world content creation.

I made this from scratch. Except the bricks. Where did I get those bricks?

I made this from scratch using in-world building tools.

10. It’s hyperlinked

On the Web, people can move easily from website to website by clicking on links, by typing in a URL, or by using a bookmark.

On the hypergrid, people can move easily from grid to grid by using a hypergate portal, by typing in the hypergrid address, or by using a landmark.

Moreover, on the hypergrid, it’s not just your presence that moves from one grid to another. Users bring their appearance with them, and can access their inventories, their friend lists, their groups, and their messages.

There is even a hypergrid-enabled currency that users can access on any grid that supports is, similar to the way you can pay through your own PayPal account at many different sites.


No other platform available today offers each of these features. High Fidelity might, some day. And other platforms could come along, as well.

But right now, the closest we have to a functioning metaverse is the OpenSim-based hypergrid.

Will it eventually be replaced by something else? It depends on how much of a critical mass it is able to build before new platforms come along.

Related Posts'

Maria Korolov

Maria Korolov is editor and publisher of Hypergrid Business. She has been a journalist for more than twenty years and has worked for the Chicago Tribune, Reuters, and Computerworld and has reported from over a dozen countries, including Russia and China. Follow me on Twitter @MariaKorolov.

39 Responses

  1.' Talla Adam says:

    I always said Opensim and Hypergrid was the nearest thing to a connected Metaverse and I even set up G+ Opensim Virtual two years ago to find out, if possible, just if there really was a cross grid community to speak of. And I think it has proved correct. And it is still growing!
    G+ Opensim Virtual community @

  2.' Adam Time says:

    What a wonderful well written article. Expanding in to space time and distance still at the point of a needle. Cosmic.

  3. Great article! Hopefully High Fidelity will be the next step in this journey. I am very excited about it since hearing Philip speak about it at OSCC this year. He even mentioned the possibility of directly porting OS instances into HF. That would be amazing!

    Maybe it will even be possible to Hypergrid back and forth to some extent. I think/hope there will always be a place of Opensim.

    •' skylifegrid says:

      Mmmm Central Grid Services run by Philips crew ? is that not like what we have now ? I hear that This High Fidelity is just the same as Opensim Hypergrid Allowing users to connect to a central database and charge people to connect to it Correct me if I am wrong I only have done a smidgen of research on it. any good literature you would recommend Ozwell Wayfarer about High Fidelity?

      • The impression I got was more along the lines of what Maria mentions here:

        “or example, in AOL, users got everything they wanted in one place search, news, shopping, friends, personals, messages, and so on.

        On the Web, they might get their search from Google, their shopping from Amazon, their classifieds on Craigslist.”

        So while Philips crew might be one provider of grid services, they wont be the only one. I imagine it wont look to dissimilar from the OS we know now, save maybe a more robust way to travel between “worlds” or “grids”.

        I am only grabbing this from the bits of info I have gathered here and there. As I said, listening to Philips 2 hour long talk at the OSCC this year was enlightening/interesting. Maybe start there.

  4.' skylifegrid says:

    I just love it all 🙂 Exciting times we live Keep it coming!

  5.' guest says:

    Good article and points all well taken, but the software simply has to get better, and a lot better, to meet these expectations.

  6. While I agree with you in part Geir, there are a few things that I just dont see OS overcoming before the “next big thing” comes along and sits on its face.

    1. Accessibility – We have been trying for ages to get some kind of working, freely accessible and stable web viewer for ages now and I dont see much progress being made in that area.

    2. Interactivity – Compared to modern game engines, LSL is positively stunted in what is possible in terms of interactivity. Again, I dont see anyone making massive strides to improve this. One of the formats biggest drawbacks *both SL and OS* is a lack of things to do. Many regions are just pretty dioramas. LL are trying to address this with the new “Experience Tools” in a way.

    Not putting what we have down, but I expect HF or one of the new VRs to do both the above substantially better than OS/SL, and I believe that accessibility + compelling experiences are the way to more of a mass market.

    I agree that the only tangible thing shown at OSCC was the avatars still, but I was more interested by the concepts he discussed, specifically being able to directly port builds from OS to HF. If I can have my old OS builds but new avis, new tools and new possibilities then why wouldnt I move?

    • There was a talk about OSSL about creating a web viewer — I really hoped that this would have built some momentum behind it. I guess not. And I had high hopes that the new grid owners association would sponsor the creation of one — and that might still happen.

      Not sure what you mean about interactivity. We have music, socializing, etc…. It’s not really a platform for games — there are better platforms for that. Just as the early Web wasn’t particularly great for games, either. In fact, a good portion of the games people play are STILL on proprietary platforms, using proprietary viewers (like World of Warcraft) or even proprietary hardware (like Playstations and Xboxes)…

      As far as what will keep people here if they can move regions over…. well, it’s the people. If all your friends are here, if all the stuff you do with them online is here, then you’ll stay where your friends are. And newcomers arriving aren’t going to go to the platform with the snazziest graphics and avatars — they’ll go to the platform with the most people.

      Cloud Party proved that a simple web viewer and interface wasn’t enough to attract users. Blue Mars proved the same for mesh.

      Will High Fidelity be enough of an improvement over OpenSim in two years or whenever it’s ready to go live to get people to move over?


      Or maybe they’ll be compatible, and we can move back and forth, which means that you can move your region over AND still have your friends — best of both worlds.

      • What I mean with regards to interactivity is not that we have none, but that its limited. Dont we hear time and again that there’s nothing to “do” in SL/OS from newbies? I am sure I dont need to expound the failings in this area point for point. I just think a “fresh start” VW has a better chance of getting this right faster than OS does. And a big injection of venture capital wont hurt its chances either.

        The metaverse in my mind is/should be just as much for game purposes as anything else. One of the biggest and longest lasting community’s in Second Life is the RP community in all its clans and groups.

        Also, I notice you dont mention mobile gaming. For that I just need a phone and I think that beats everything these days by a country mile. But that opens a whole other can of worms. We probably need to be mobile for the next gen, which probably means stepping BACK in terms of graphics and eye candy compered to SL. At least initially until mobile processor power comes close to a modern desktop. So if HF is going for mobile and low latency, I would not expect those avis to get much prettier before release. You cant have it all.

        Its not easy right now for someone to come in and create a “guided experience” like in a game but not limited to that context. Or how about the Myriad RP system which has been forever in the making. I am not saying these things should be super easy, but creating a basic game experience should not take months or years.

        Ultimately no single factor is the magic key.

        “Or maybe they’ll be compatible, and we can move back and forth, which means that you can move your region over AND still have your friends — best of both worlds.” – Thats exactly why I found Philips comment exciting and I think it would be a smart move for early adoption to do just that.

        •' Talla Adam says:

          Sorry Ozwell I don’t agree that we have no interactivity. Nara Malone and the writers of Greyville have produced a number if Interactive Fiction games to play and there are others as well as a hud available on Kitely market to create most types of adventure, hunts, etc. There is also a lot of work being done with NPC’s to support gaming. Add to that the improving state of Bullet physics. Have you looked at one Ingen Lab videos lately? There are some very useful weapons now available on Kitely market too and sailing boats for those who enjoy sailing and surfing, etc. Check out Mike Hart’s work with Next reality. He is the motor king! and he has some pretty able NPC’s doing some serves as well. And he is not the only one because there are some great rides around too like roller coaster and trains.

          You also mentioned roleplay and the Myraid RP system, which is a fully working combat HUD for Opensim for which weapons are available. Opensim could serve the RP community well if more role players came over and did some cross-grid stories between the many grids and even Second Life. The potential is there – not the lack of. What we lack is enough people to fill the roles and do the stories and play the games. But, really, it is getting better, just not quick enough for everyone. But there are small groups working on it in different grids. Please don’t think no one is trying. They are, I promise you and 2015 will see a lot more happening. Heck, I’m working on stuff myself to and scripting with LSL, OSSL which is not so limited. I can do stuff with php too additionally to support special functions as well so it is just a matter of time in my view. HF will arrive, sure but I’m not waiting for it. There is work to do now with what we have. Heck, we can make stuff happen and for sure we are trying!

          • “What I mean with regards to interactivity is not that we have none, but that its limited.” – Ozwell Wayfarer directly above 😛

            I am not trying to say, nor did I actually ever say at all that no interactivity exists, nor am I saying that there are not people out there making an excellent effort with the tools we have. Sorry if that comes off a little strong, but I want to be emphatic here.

            I never said we had no interactivity.

            I am a bit of a game head, so I have experienced a lot of the cooler UI and game mechanics of recent roleplay games. Its hard to describe these things if you dont have direct experience of them. Sorry if you do but I will assume for conversation sake that you dont.

            Say you pick up a book. Now in SL, it might be scripted to open, it will even maybe give you a notecard with a story on it but in Skyrim or Dragon Age, pick up a book and you can flick through it page by page. There are a million potential improvements I could name with “land” alone without going into mechanics at all. Splat Maps for terrain, procedural vegitation. Multi-level water.

            Last I was aware, myriad was at “Preview 5 release” implying to me that it is still incomplete. I a lot of it does indeed work, but not all so far as my understanding of the release notes goes.

            But what I AM saying is we need better, more powerful, more intuitive tools for creating experiences. LL recognize this and are trying to do something in that regard with the “experience project”.

            I have always been a proponent of the philosophy that Opensource AND commercial venture are the key to success, which is another reason I think HF will be a winner as it is basing its whole business model on that exact mix. I am not saying all these things are not possible in OS. I just think they are improbable to happen before someone else does it.

          •' Talla Adam says:

            Okay, I jumped a bit there. My bad! But you did come over in a very dismissive manner like Opensim was lacking so much. I am not a serious video game player, that is true but I have looked over the shoulders of our youngone’s to see what they play (making sure it’s safe clean fun, your know) and I get all that what you were saying about the game experience but niether Second Life nor Opensim were made for video gaming really. But they are great for role play and interactive games of adventure and such. Myraid is fully functional as a combat meter. It is, in my view, a bit bloated with adventuring stuff too and things the coder wants in it. It can be used just as a basic meter though. Besides that I have my own combat meter anyway which is a development of the one I coded for Second Life and it uses the server side web data the SL meter uses so players could use their same ID in either SL or Opensim. So, a lot of things are possible.

            I even have a scripted book that opens, or I did in my OSG inventory but I have the script back-up somewhere, lol. I just don’t think we are so limited and you only have to take a trip to Fred’s Outworldz grid to see some truly amazing works of scripted NPC’s and other stuff to get the idea. Fred has even said there are some things in OS now that just can’t be done in SL.

            I would like to see new things for OS, for sure. Yes, I would like to see the use of voxels and a new water surface like that I have seen for Unity. There is always more that can be done but my argument is we are not that limited any more which you seem to take the view we are, Ozwell. OS has come on leaps and bounds and while I too am interested in HF it is way off yet and we may yet get some more tools to work with before too long.

          • I did not mean to sound dismissive. In fact I did try in both posts to point out that I was not dismissing what we have, but rather looking to the future.

            In some ways its kind of like trying to describe Windows to someone who has only ever seen DOS (that sounds a little snobby too, sorry!) I know we have things like those I describe right now. Say the opening book. I have seen plenty of opening books in SL/OS, I have however as yet to see a book that when I pick it up, fills my whole screen like a HUD and has interactive pages I can scroll through to read a story, or animated pictures, or narration. Or any number of things I know to be possible from playing games. Obviously, functionality like that has possibilities outside of gaming. The reason I draw on gaming analogies is because games are the thing out there most like virtual worlds. But many of the things now possible that would really help virtual worlds are not possible not because of a lack of talented people, but because the framework to make it work is too cumbersome or does not exist.

            But just to be clear what Im trying to say here: I know all this stuff is possible in OS, I just dont think its probable. The codebase is old, LSL is limited compared to other scripting languages like Java and C++ and though I know OS can grow and evolve, the issue is pace. I think something will eventually outpace it.

            I am sorry if that seems a little pessimistic, but I just feel thats a realistic assumption. The scale of the task is just too big. But I hope for cross-compatibility, so we dont have to make the choice.

      •' Geir Nklebye says:

        > about creating a web viewer — I really hoped that this would have built some momentum behind it. I guess not. And I had high hopes that the new grid owners association would sponsor the creation of one —

        There are really two issues to this:

        To create a web viewer you basically have to dump a massive viewer plugin into the viewer that needs to tap on the GPU in the machine the viewer is running in, so what you have is a viewer + the memory and processing overhead in having it exist inside a web browser. Alternatively you have to rewrite the backend (OS Server) to the extent it can sensibly respond to browser and touch events and break the floaters out of the current viewer to exist as widgets in the web interface. By this time you are a long way to SL v2


        You can attempt to render the current scene view on a remote server and stream the result to the device in a browser plugin. This can work in small scale deployments, but the remote server needs the same GPU resources to render the scene and code it into a video stream as a PC with a viewer does. A server can be run somewhat more efficient than a desktop in that it can take clients 24/7 and a single client session often does not take more than up to an hour.

        However, having a farm of servers with powerful GPUs at the datacenter both creates a cooling issue and a power issue that drives up cost of the installation. It also creates a packing issue because most datacenter blades cannot accommodate the GPU cards required so you need to move to much bulkier boxes. This drives up the cost of the solution.

        Equally important for the streaming solution is the bandwidth requirement issue as seen bot from the datacenter side and not the least from the client side.

        It is often said yes, but people will only use it on wi-fi. This statement makes the assumption that everyone has access to wi-fi and in Western Europe and the US/Canada that might be true, but…

        In Eastern Europe, Asia, Africa and other locations of the world the populations simply did not have phones so there is no wired connection to the households or even businesses. TV is not distributed on cable, but via arial or satellite so there is no cable infrastructure to run data across. The telcos in these countries have tactically brought phone and rudimentary internet connections to their populations by building and deploying mobile networks, but the ability to carry large amounts of streamed content in these networks are close to non existent. So in these geos such a solution not viable.

        There is one further aspect of this, and that is one that Apple discovered when they brought out the iPhone. The original model for deploying apps to the users were through web-apps, but people did not like that at all. It was only after Steve Jobs more or less were forced to make an SDK to create discrete apps that it all took off.

        It came as a massive surprise both to them and the industry in general that people preferred the app model because it made it easier for them to break down tasks in discrete units, and it made them focus better than trying to run everything in the browser as Microsoft was the main proponent of in those days. Because of this finding I am now of the opinion running OpenSim /SL in a browser session is a flawed model. (I was also all for running as much as possible in the browser back then.)

        As for funding see a separate reply

      •' Geir Nklebye says:

        > I had high hopes that the new grid owners association would sponsor the creation of…

        For anyone to sponsor anything but minor development a roadmap for the OpenSim product must be created and agreed to. As far as I know, at this point in time, not even a rudimentary development roadmap exist.

        I have tried to figure out how decisions on how the development work is driven forward, and my best assessment is that strong interest by individuals amongst the developers drives it forward in leaps and bounds, and often not necessarily in a direction that is overall beneficial. Because these individuals change over time, issues that had high focus in one period is defocused or neglected in other periods. This leaves the product over-developed in places, while long standing bugs and issues are never fixed or updated to balance other development.

        In my judgement, to attract major funding for development, OpenSim needs to be forked after a roadmap with clear development goals has been made. This possibly needs to encompass both the current viewer and the development of a mobile viewer.

        It is close to impossible to drive progress to a roadmap forward with the speed required to make a difference in the current development model based on the voluntary work of individuals.

        •' guest says:

          Absolutely right on! If you take a look at the opensimulator manta stats, you will see that while there are a reasonable number of contributors, only a handful of individuals are making most of the changes. The project looks to me more like hobby coding than true open source. As well, the architecture is suspect. I strongly question anything open source done in C# rather than Java.

          •' Geir Nklebye says:

            To call it hobby coding is perhaps a bit of a stretch, but what is apparent is that the current development effort does not deliver sufficient improvements and reliability that can be used in a serious business setting without adding up to significant coding resources to an implementation.

            One of the things that surprised me most about the OSgrid crash was the revelation that the test grid had not used the asset server shipped in the release product, and thereby not tested or hardened it sufficiently, but rather used a 3rd party module that so far has yet to produce a recoverable system.

            Only a few days ago it was said in the weekly developer meeting (minutes) or on the dev mailing list (can’t remember which right now) that the built in asset server is not scalable. I’d say that the asset server is the Achilles heel of any opensim grid implementation, and should get all the love and care it needs to shine.

          •' Geir Nklebye says:

            Hi and Happy New Year! Looks like you did some good work there. And I can attest to worlds and assets loading fast when transferring items to Kitely from xmir.

            I looked at the bit on http texture loading in one of the posts and I have a vague recollection JustinCC tweeted something about how that could be enabled in standard OpenSim too. I’ll see if I can find it and actually understand what he said so I can test it.

            I believe also Singularity now has http texture support for OpenSim.

            I have experienced very good performance increase using PostgreSQL for the asset server, but the adapter is still a bit buggy, but getting there… 🙂 Autovacuum will also mop up many duplicates, and kick unreferenced records out of the database to keep it leaner.

          •' Ilan Tochner says:

            There’s a lot more to it than just turning on HTTP textures support (which all V2/V3 version viewers support). One key difference being that HTTP handling is not done by OpenSim in Kitely. In other words, the work required to transfer those assets in Kitely is offloaded to different servers instead of being handled by the OpenSim simulator itself.

            Improving Inventory and Assets requires a lot more than just changing how the data is persisted. Switching from MySQL to PostgreSQL can help but it isn’t enough if you want to have good performance in a grid setting where everything isn’t done inside the same server. Try comparing how much time it takes to move large inventory folders in Kitely verses how long it takes in other grids that don’t have the benefit of our Inventory System.

          •' Cookie Monster says:

            say the guy whos sims are always offline and take 5 minutes to 5 hours for someone to teleport to a sim on kitely then another 5 to 30 minutes for everything on that sim to fully load.
            I rather take my chances with a always on grid like metro.

          •' lmpierce says:

            I wonder about your figures for the time it takes to load sims in Kitely. I realize you’re using hyperbole to express a critique of the Kitely on-demand system, but still, it helps to keep the values realistic.

            As a test, I loaded the Kitely world Serenity Island – Prim Capability Test. The prim count is 104,481 (plus hundreds of script instances and many high resolution textures (512 and 1024)). The total time from launch to entrance into the world was one minute, 26 seconds. After teleportation into the world, local objects fully resolved by 12 seconds later. From that point, walking around the world worked well, and all prims within view were visible (Draw distance = 128m). Use HTTP for receiving textures was checked ‘on’.

            This is noticeably better than when Kitely began as they have improved the world loading times over time. Since Kitely worlds have a limit of 100,000 prims, the test I performed indicates that accessing any Kitely world is faster than the estimates you have stated, even at the low end.

          •' Ilan Tochner says:

            Most people spend less than 15 seconds in a Kitely Transfer station when logging into a world which wasn’t already active. Once they login into the world the time it will take for it to rezz will almost always be limited by their own download speed and not by the speed of the sim itself.

            If you want to spread FUD then I suggest you find another angle. 🙂

          •' Geir Nklebye says:

            The link I gave describes how you can offload the texture loading to another server(s load balanced) so the simulator is not used for this at all. I’ll make a test run of that scenario today and see how it works.

            What do you consider a large inventory folder? I would not think that moving a folder in the Inventory should have to update more than max two database records, but what do I know, haha.

          •' Ilan Tochner says:

            Rezzing objects requires downloading both the Assets and their metadata. Viewer requests for those things are (or at least were, we haven’t tested the latest SL viewer releases) very inefficient. It isn’t enough to have textures offloaded, you need to optimize the way object information is transfered within the OpenSim architecture. As stated previously, the persistence layer implementation is just one of the bottlenecks.

            People have folders with thousands of items, try moving one such folder on some OpenSim grid other than Kitely and see how long it takes. Moving an inventory folder is extremely inneficent in OpenSim, especially when it uses ROBUST. The process involves a lot more than two database queries / updates. We had to do a lot of optimizations to improve Inventory handling.

          •' Geir Nklebye says:

            OK, I reconfigured, so textures are now served up http from a different machine in a Robust instance doing only that. Works pretty good (so far) it seems.

            I am not quite with you on the overhead of moving large folders. I have a folder of about 1000 items and moving it in inventory is instantaneous. It takes less than 10 database transaction (I could not measure it more accurately.) This is on Robust with the more or less bleeding edge dev code.

          •' Ilan Tochner says:

            Is the sim running on a different physical server than the inventory server? Is this on a live system where other people are using the system as well? What operating system is the server running on? What database is it using?

            When we created our asset and inventory systems in 2011 the ones provided with OpenSim were very unoptimized. When we last updated to the OpenSim 0.8 release, things were still problematic. They may have made changes to these systems for the OpenSim Community Conference 2014 (we haven’t checked so I can’t comment on the current status of the dev branch).

          •' Geir Nklebye says:

            – Is the sim running on a different physical server than the inventory server?

            Partly – main Robust is running on the same server as the simulator. The Robust that is serving up the textures via http is running on the same server as the database

            – Is this on a live system where other people are using the system as well?

            Yes – hg: grid dot xmir dot org colon 8002 (any post with a link makes the post go to moderation, so truncate to a proper address)

            – What operating system is the server running on?

            OS X 10.10.1 with OS X Server 4.0 with mono for the simulator and OS X 9.5.4 with OS X Server 3.2 and mono for the database / http texture server

            – What database is it using?

            PostgreSQL 9.3.4

            Many of the changes made for the OSCC has probably caused more issues than they have solved, so I am running that code with the throttles disabled in addition to tuning some other network params for opensim to be in line with operating system capabilities.

          •' Ilan Tochner says:

            In a live deployment that needs to scale you won’t have the sim sharing the same server as the main ROBUST, I therefore suggest you test this when they are located on different servers. Also test when the viewer is not located on the same network as the servers it is connecting to.

            I believe the people who developed the PostgreSQL OpenSim connector mentioned optimizing some of the database queries. IIRC, that would help to some regard but unless they also optimized the communication protocol between the sim, the viewer, and ROBUST then there will be bottlenecks when OpenSim is set up and used in a realistic network configuration.

          •' Geir Nklebye says:

            I absolutely agree on the not sharing robust on a simulator server, and it will be cleaned up once the database server is upgraded to OS X 10.10 🙂

            I suspect it is the PGSQL adapter being a little bit more intelligent about how it drives the database. On OAR loads I have consistently seen 45-50% better performance than MySQL on the same hardware. On database backups it delivers close to 120 Mbyte/sec sustained, meaning it drives the gigabit ethernet to saturation, which is not bad. MySQL was not even close to deliver the same performance.

            As mentioned above, tuning network params and turning off throttles has virtually eliminated the

            [LOGHTTP]: Slow SynchronousRestObject request 11881 GET to …

            type messages we have all seen on the console both when the user is local or arrives via hypergrid.

            I also let the simulators have all the threads they can use because I found that mono was significantly thread-starved in certain situations with the then default settings. JustinCC has since upped the default settings for some of the pools significantly.

    •' Geir Nklebye says:

      See my answer to Maria below on the web viewer.

      > If I can have my old OS builds but new avis, new tools and new possibilities then why wouldn’t I move?

      This is possibly a pipe dream as they would have to recreate the functionality to port your old OS builds (with scripting) in the new HF product, weighing it down.

      Besides, the waiting time can be spent on upgrading the current product to match without breaking current content. Further, and equally important, is that SL v2 will be highly disruptive and all the current residents may be looking for a heaven they can move to when LL starts to pull the plug on them.

      • Well, as I mentioned one of the reasons I would personally consider moving is to take advantage of better/more powerful/more intuitive scripting and opportunities creating compelling single user experiences. So I am not so concerned on the scripting aspect. Just the physical assets. I would not expect them to do backwards compatibility with LSL. Even LL are not doing that with SL2.

        The whole point of “upgrading the current product to match without breaking current content” is exactly why I think eventually something will come along to beat SL/OS.

        Bottom line is SL and OS are old tech, and realistically, regardless of the impressive progress OS has made I dont think the waiting time between now and the next gen will be enough for either to come up to the same standard. I dont even think it can practically be done. Otherwise LL would be doing it with SL. There comes a point where upgrading just isnt an option anymore. Otherwise we would all still be using the first Mobile Phones or Computers we ever bought (provided they didnt break).

        Its a nice idea that OS can go on evolving forever and I hope it does for a long time yet, but there comes a point where it wont be possible/practical. Thats not to say there wont be a place OS. Utherverse is still around after all (…..I think).

        •' Geir Nklebye says:

          I think I fundamentally disagree on the ability to upgrade a system to significantly evolve over a long time and still stay at the cutting edge. 🙂

          Having used and worked on Mac OS through multiple processor generations from the Motorola 68k series, through PowerPC, to Intel 32 and later 64 bit, and lately ARM 32 and 64 bit without virtually no loss of content or even software, my perspective on what is possible is perhaps different.

          My current system image, on the machine I type this, was originally installed as Mac OS 8.6 in 1998 and has been upgraded unbroken through to Mac OS 9.2 from which is was upgraded to OS X 10.1 in 2001, and further upgraded to 10.5 in 2007 all the time on a PowerPC processor. It was switched over to an Intel processor on OS X 10.6 in 2009 at the time much of the software was running in PowerPC emulation on the Intel processor, gradually moving to a clean Intel 64 bit family of applications (the SL viewer is an exception.) The system is now running 10.10.1 (for reference Mac OS 1.0 was released in 1984.)

          I have word processing and Excel documents created back in 1992 on Mac OS 7.1 on a Motorola 68040 processor on my hard drive that I still can open, including embedded graphics.

          You can say that OS /SL is hopelessly architectured, but then again Hypergrid provides for the horizontal scaling SL lacks. Hypergrid also provides for the horizontal cultural scaling SL lacks and which I consider the more important. Although the current OS server in many ways are sub standard, it is possible to fix it with a reasonable amount of effort.

          • “Having used and worked on Mac OS through multiple processor generations from the Motorola 68k series, through PowerPC, to Intel 32 and later 64 bit, and lately ARM 32 and 64 bit without virtually no loss of content or even software, my perspective on what is possible is perhaps different.” – …….Well, thats certainly impressive, but I think it would be fairly safe to say you are in a minority.

            You are obviously far more into the technical aspect of this than I am. I dont understand your final statement regarding HG at all. So far as I understand HG is tied to OS and its derivatives. If you mean different grids/regions can rund different versions of OS, that happens in SL too to a limited extent. So I dont see how that is helpful. Sorry if I have the wrong end of the stick there.

            “Although the current OS server in many ways are sub standard, it is possible to fix it with a reasonable amount of effort.” – Thats what I am saying. I know its possible. Just improbable. I think a fresh start in some ways is much better than trying to plug the myriad of holes in a multitude of areas without breaking anything in the process.

            Your technical aptitude in this area might also be why you dont perceive the issues I perceive to be such big problems. I am coming at this from “average joe who knows how to start a PC but not much else” perspective.

          •' Geir Nklebye says:

            SecondLife can become a member of the Hypergrid by more or less enabling the hyper gird protocols in their servers and publish gateway addresses.

            Although the internal service delivery model in SecondLife has drifted somewhat, it is fundamentally the same as for, say Kitely, and the sims and accounts belonging to that grid.

            The biggest difference is the amount of data (300+ TB), registered accounts (24+ million), the size of the economy, and concurrency which for SL peaks at around 60k every day. In comparison I don’t know of any OpenSim grid that has concurrency over 1000 users.

            The centralized service delivery model for SL represents a scalability problem, particularly when it comes to deliver data to the sims and the users. Handling a 300+ TB database is also not non-trivial in terms of backup, performance, availability and security.

            Hypergrid with OpenSim does not have the same scalability issue in that you can spread the load across multiple instances, multiple (asset) databases, but equally important you can distribute the sims much closer to the users. Because of the way the system is put together latency on the data communication between the viewer and the simulator becomes an issue. With the SL servers all located in the US, I will always (being in Europe) get a quite substantial delay on the line making movements more choppy and imprecise for instance. If I were on a mobile connection the choppiness would be even more pronounced.

            Take Germangrid as an example; they primarily have German users and with the servers (presumably) located in Germany they will get better performance than if they were located overseas.

            Equally important is that the servers are operating under German legislation meaning they have to observe the privacy and consumer legislation that is mandated in the EU, and which is very different from the US. As an example, the SL TOS is simply invalidated by any European court because they don’t satisfy the minimum requirements of the consumer legislation. So our consumer authorities says we can simply disregard them. Hypergrid therefore has the potential of regulating the legal environment better for the users. Visiting a grid in a different legislation is like going on vacation to a different country.

            By the same token, Hypergrid opens for larger cultural diversity and more content, simply because certain things like online gambling may be legal in some legislations and if the grid is located such, it can offer it, whereas there may be other restrictions that apply for the same grid.

            It can also offer users content and guidelines for accessing content more in line with the cultural norms of the location the user resides in. We saw that as an example in SL in 2009 where Europeans left in hordes over the age verification guidelines which asked for information LL never was licensed to hold according to EU privacy legislation, but also because they imposed some standards which many found both prudish and strange compared to their local norms.

            This is not special to LL, but we have also seen it with Apple and the App store where there has been similar clashes between Californian legislation /PC and European content providers. We also see it with Google where the EU now really want the company split, and where there has been major issues about the subject “the right to be forgotten” where Google has been forced to offer solutions to have information about an individual removed from their search indexes.

            The distributed nature of Hypergrid has many advantages, but there are still some choppiness in how members (grids) communicate with each other, search is largely missing and communication outside a grid is not slick. These are, however, issues that can be addressed and resolved.

            Hi-Fi (as I understand it) will also be a (highly) distributed system, and they will face exactly the same type of issues during their implementation as OpenSim currently does. One can expect there will be significant delay between the initial offering and these issues are resolved, so also their users will experience a bumpy ride in the beginning.

  7.' Han Held says:

    >I see many of you hail High Fidelity as the NEXT BIG THING, but I
    thought the presentation at the OSCC was exceptionally weak where the
    main feature was the ability to display emotions on the avatar’s face.
    The rest was conceptually a non-event.

    There’s a huge difference -Opensim and Secondlife are both hostile to tablets and general purpose devices. That’s in the process of changing but there’s no guarantee that it will successfully change.

    My understanding is that Hifi is being developed from the ground up to run on and take advantage of smaller devices.

    •' Geir Nklebye says:

      I agree with you on that assessment. LL is running a massive risk there. Which is an opportunity for current OpenSim. 🙂

      If you look at the state of the Mac viewer, it just shows how far back they are in the development process. It is built with frameworks Apple discontinued 4 major operating system versions back (it is like building for a Windows 2000 target.) If they, at all, want to be on iOS it must be rewritten ground up in Objective C or Swift + Metal. Starting with the Mac viewer would be a good training experience, but they are not even there yet…

  8.' Geir Nklebye says:

    > Perhaps the time has come to take a long hard look at basic open simulator server architecture.

    I can agree on that but specifically the asset server issue has two aspects:

    1. The asset server (with database adapters) in the shipped version needs to be tested through and through for performance, security and transaction integrity, and the development team needs to know how far it scales so everyone can make informed decisions on when to use it and when to move to

    2. A more scalable asset server that can be deployed for larger grids. There needs to be tested alternatives (as in supported) for this also.

    In general it is important that opensim can sensibly run both on “sim on a stick” type configurations to large professional grids