Justin recently wrote an article about the likelihood of the concept of a “Grid” to vanish fairly completely. I think he’s bang on there and I expect to see things play out fairly similar to how he describes. The reason for this is that the concept of a “Grid” is completely and utterly irrelevant in the long term.

What?

I suspect in the long term, some of the models presented by alternate virtual worlds (Croquet in particular) are largely correct. While the ability to “load balance” a larger 3D space across multiple servers by partitioning the geometry accordingly is a very valid feature - it restricts you to creating giant contiguous landmasses.

And I dont think this is something either users nor companies want.

The analogue with the traditional web is the concept of somewhere like Geocities - under the contiguous space model, every user from geocities has their webspace crammed right next to someone elses, and you can see it whether you like it or not.

If someone makes any parallels here with the Second Life™ Mainland, you are probably right on target - it’s probably one of the reasons that Private Islands in Second Life eclipse the number of mainland regions. Now that’s not to say that users wont want to congregate together on occasion - consider the Steampunk themed Caledon sims - but in that occasion it is strictly by choice, and not representative of the majority of users.

Supporting both is of course a priority - but I suspect in the long term that the abitrary collections of regions wont be crammed together. Most will be linked by the same technologies that link the internet today - IP and DNS, and any organisation will be built ontop of that rather than the concept of the grid itself.

So what about users?

Right now - the single most inconvenient factor to visiting the OpenSim grids today is the requirement that you create a user account before visiting. Unlike email where you can login with a single username and send a message anywhere - you need a seperate account for each server you want to visit.

If we seperate these out (as the AWG OGP spec does) we get to the point where your username comes from someone like an email provider (ISP, Free Hosting site, etc), and the regions are seperate things that you can connect to like visiting a webpage.

In this case, grids become groups of commonly themed regions that are visitable with either commonly themed URLs (ogp://grid.com/regionname/x/y/z/) or contiguous landmasses and not much more.

One of the beauties of the internet’s design is that you only need a single number to represent every server connected (an IP address), there’s millions of servers connected each with their own address - if you tried to organize those millions of servers into a set of finite artificial constructs, you would probably fail - the operators of those servers tend to like to run their own environments and not be reliant on other people for stability and uptime (there’s a bit of a commercial incentive there).

Why proposing things that rely on grids is probably a bad idea

There’s been a lot of suggestions lately about things like content enforcement being locked to a specific grid for example. The catch here is that there’s potentially one “grid” for every independent region online under the AWG spec. Only places such as the Caledon-equivilents are forming grids with multiple servers in them.

In this case the question becomes - if grids are not a good analogue for the operator group, what is? The answer here is probably the hosting companies. While I don’t have a firm number here - I’d say that probably 50-80% of the web hosting on the internet today is done by a small group of companies and their resellers (1and1, GoDaddy, etc) - and those are the groups you will want to get contracts for enforcement with.

The remainder may sign onto the contracts, but you can easily get the large groups with a smaller amount of effort just by hitting the hosting companies.

It’s a simple enough question - often when I see a company or group get involved in OpenSim, the first thing that springs to their mind is “Well, we must launch our own grid!”, with each grid having it’s own isolated set of users, inventory, assets and regions the question must be asked: “Why?”.

But in most cases, this often isn’t the best of ideas, and short of internal uses it’s often counterproductive duplicating effort others have undertaken already. Running a grid properly takes a lot of work - especially if you are doing frequent updates, if your intent is on making it a popular destination for random visitors, then you have extra work publicizing and convincing people to create an account and login.

All in all, it’s a lot of work for relatively little benefit.

Alternatives to starting your own grid

Attach your region[s] to an existing grid - I personally like recommending OSGrid.org - one of the downsides to this approach is that you are relying on the grid provider for your infrastructure - if they either neglect to update frequently you may run into problems, likewise if they have downtime or stability problems, you are tethered to them good and bad.

So, why do it then?

The short answer is to do it for your users, if you require users to register an account on your grid to come view your trinket, you have effectively annihilated any chance of instant gratification so to speak. The difference between opening a map on a popular grid, and logging out then relogging onto your grid is huge. It requires a good deal of effort to to so.

The longer answer lies in pooling resources - in this case users, users may log in to view your trinket, but after they are done with it, go visit someone elses trinket nearby, the key here is that people who come for other peoples shinies may end up visiting yours too - there’s no zero-sum game here, and offering a multitude of attractions is more powerful a drawcard than yours alone.

When you should make your own grid?

There’s plenty of reasons to run your own grid however - the biggest one is privacy and wanting to run something entirely behind a single corporate firewall. Protecting intellectual property and trade secrets requires making reasonable protections when possible, and public grids are not well suited for this.

Another reason is when you want to run something solely independently - such as say a specific concert or event, where you know that your demands require customisations on the grid software itself. In these cases unless you are able to make an arrangement with the grid operator you may run into problems and hosting oneself is a good idea.

So which grids allow outside regions to connect?

OSGrid.org is the best for this since the entire grid is hosted externally - making sure this works smoothly is a goal of the osgrid admins. Connecting a region is free and there are no real terms of service at the moment to speak of, other than “don’t break the law.”

DeepGrid has allowed regions to connect since the beggining - although the grid server software is updated less frequently than the OSGrid ones. A common terms of service do apply for this and region spaces need to be reserved via the website first before connecting (rather than first-come-first-served, price is still free).

CentralGrid offer region connections too however connecting a region invokes a large fee that isnt charged anywhere else, for as far as I can see no advantage. I’m not impressed, but you may be.