The Object Network: comparison to other virtual world systems
There have been numerous attempts to create open Metaverses, either from scratch or through joining up smaller or separate virtual worlds. There is little agreement on what it even means to be a Metaverse.
Most 3D virtual worlds have the basic model of avatars, people or players, worlds or regions and objects or items. Mnemonically: People, Places and Property. So a Metaverse will at least have those elements.
The Object Network has a vision of the Metaverse that includes the following features:
The Object Network meets these requirements through the power of the Link.
Can your system meet the Linkiverse Challenge...?

The Linkiverse Challenge

I build a nice garden. You do the same. We link our gardens together into a continuous landscape with a link in each direction, setting appropriate mutual permissions. I see that your garden has a tree in it whose leaves I like, so I walk across and grab the link to one of the leaves, then walk back to my own garden and swap all the leaves of my tree to this one. It's still your tree and the leaves can only be changed by you, so if you change the leaves to red in autumn, mine change too. A bird flies from your tree to mine.
Note that this scenario should work across independently-owned hosts or devices.

Open and Closed Virtual Worlds

Consider three broad categories of popular Virtual World:
These worlds and games are generally centralised, closed, walled garden, proprietary and incompatible, with isolated regions, rooms or worlds even within their own servers.
There are many examples of more open source, more open standards-based virtual worlds: Being open is a significant improvement over the proprietary offerings above, as it at least offers hope of delivering the single global Metaverse, in the same manner as the Web swept the world, wiping out the walled gardens of Compuserve and AOL.

Joining services and worlds

It is of course quite a challenge to sew any of the closed systems together. Even to allow teleporting and property transfer between the isolated regions and worlds within each is challenging, never mind between.
The open systems offer a little more hope, and some have started to allow this. However, it's still jarring or clunky to move or teleport across worlds, regions or rooms. The attempts to allow interoperability between each of these worlds is definitely an early work-in-progress.
There follows a quick summary of the ability of each of the above open systems to meet the Linkiverse Challenge within their own offering not between them in any way. Note that Minetest isn't attempting to take a place in the Metaverse, last time I searched, so is excluded.

Traditional

OpenSimulator is quite an old system based on Second Life. You may download the Second Life or Firestorm client and connect to a service such as the OSGrid or Kitely, which run the OpenSimulator servers. Each such server offering has its own accounts and 2D mapped fixed grid, which allows seamless walking between regions owned by different people, as long as permitted. Teleporting between regions on different grid services requires "HyperGrid" support, which is similar to URL-jumping in Web-based systems. It seems you can carry stuff along on this journey, and pick up objects to bring home.
It's almost there in terms of the challenge above, but it has a very heavy, clunky feel (for people brought up on Minecraft's super-slick experience!) and all of the associated websites have extremely dated UX. The HyperGrid tech is bolted on to the once-proprietary Second Life protocol so overall it's probably not going to be a pretty technology. All of which doesn't inspire hope in the platform to form the basis of a future Metaverse.
Vircadia is an open source fork of the abandoned "High Fidelity" initiative. It has a commercial aspect and they currently seem to be moving the project more into the Web category. Overte is a fork of Vircadia that intends to remain non-profit and apparently non-Web. First impressions of Vircadia are of early development of a heavyweight stack, reflected in the size, crashability, slowness and general user hostility of the download-and-run experience. Apparently it is possible to travel between worlds, within the Vircadia universe, though, but without being able to easily try it, one can only assume it's as clunky as OpenSimulator. Overte has a much more successful first experience and you can quickly get exploring places, creating objects, etc. It has the teleport jump for going between worlds, but you can't seem to carry things with you.

Web-based in general

Hubs, Third Room, Croquet and the blockchain-based worlds are Web-based, which comes with a set of general challenges when repurposing it to the Metaverse, because the Web's architectural model stretches beyond its comfort zone when implementing it: To get anything interesting done on the Web, you always have to drop down from the declarative model of fetching HTML and CSS to imperative Javascript and its APIs such as the low-level websockets, thus losing many of the benefits of the base Web architecture.
On a more trivial level, the fans spin up alarmingly while a page is open with 3D content, which a 2D page hardly ever does. Layering a Metaverse experience on top of a browser will inevitably use more resources, in a way it's not naturally architected for.
In summary, trying to implement the Metaverse over the Web is simply hijacking the browser and has little to do with the original Web's architecture. On the plus side, you do of course get the benefit of not needing to download an app.
In Web-based worlds, you can usually link rooms or regions together via their URLs. However, that doesn't make them seamless or contiguous; you have to do a heavy reload to the new region's entire page rather than seamlessly walking.
Property should be transferrable between worlds simply through their URLs, carried on your avatar then placed anywhere that offers permission to do so. But it doesn't seem the above systems offer this.

Non-blockchain Web-based

All the Web-based approaches are easy to get into (just go to the URL) and all follow the same model: Places being a room, region or mini world defined by a static mesh (probably glTF), then People or avatars and Property items scattered within.
Mozilla Hubs has a way to import and share media Property items and their URLs through drag'n'drop. It's really focused on bringing the 2D world into 3D, useful for business meetings and so-on: a 3D version of Zoom and Docs perhaps.
Third Room is based on Matrix; you can log in with your Matrix account. It apparently has links to our three types of content - Places or rooms, People or members, Property or entities. But if you pick up a crate entity in one room you can't walk into another room and carry it with you.
Croquet is quite unique in its approach: each Place - room, region or world - on a given URL loads up a scene model for People and Property (including bots) that is then kept in lockstep sync with all other copies through a semi-peer-to-peer message bouncer called a "reflector". The models can run physics and other behaviours locally on all browsers. They have a cute implementation of the portal concept where it pre-loads the linked site in an iframe in 3D! You won't be able to roll a ball across though, or presumably even pick up one of their "cards" (in-world Property items) and plonk it in the other site.

Blockchain Web-based

The blockchain-based worlds do allow you to walk between regions owned by different people around the world. However, their land is designed to be owned centrally overall and parcels issued centrally, and is artificially scarce in order to generate income and for speculation by landowners on the underlying cryptocurrency. These systems have separate layers for static land allocation and asset storage and dynamic functions such as presence (avatar pose, chat) and multiplayer world state. The multiplayer services allow people to make changes to or interact with the assets in-world and for item changes to be seen by everyone.
Decentraland appears to only have Places and People, not Property, other than wearables for People. So you won't be able to pick up an item in one region and drop it in another. See also User Inventory and Sync Scene State in their docs, where they offer some sideways solutions to shared state of in-world objects.
Cryptovoxels lets you change assets in-world, either permanently back to the asset server or, in a similar sideways bolt-on to Decentraland, temporarily in a centralised service called, confusingly, the Grid. The Grid lets others see the changes (e.g., doors opening), but it will revert to the static state when the region has nobody in it, and new arrivals see the static state, not the latest state. Each of these sideways world-state solutions are per-region, so it would be even harder to arrange transfer of any asset or item between separately-owned regions.
Webaverse is early in development and has little to show at the time of writing, but they promise in-world editing, an inventory and transferrable Property through NFTs.

Summary

Proprietary and commercial offerings can't deliver the "3D Web" in the same way that they couldn't have delivered the 2D one. Open offerings seem to be heading in the right direction, but each comes with substantial limitations, especially when seen in the light of our - perfectly reasonable - Linkiverse Challenge.
OpenSimulator is aged and clunky, with ex-proprietary and bolted-on protocols. Vircadia and Overte, coming from a similar technology source, seem to have inherited the same frictions, but Overte may be one to watch.
The 2D Web itself isn't quite up to transitioning to 3D without ugly bolt-ons: you have to build an entire new imperative stack of Javascript and Web APIs on top of the base architecture to support the needs of a more dynamic and data-oriented application, which ends up losing many of the benefits of the Web's architecture. It's fighting the technology, which never ends well.
Blockchain-based worlds, running in Web browsers, are not just inheriting the issues there, but add additional issues around centralisation and artificial scarcity of the underlying currencies and disjoint, centralised servers for dynamic aspects.
All the above would find our Linkiverse Challenge too difficult, at least currently.

The Object Network Metaverse

We need to start again with an open technology stack for the Metaverse - the "3D Web".
We can keep many things from the Web, in particular links or URLs and the cacheing and shared data types that they imply.
But we need to move on from the Web's "user second class, static document" model, to a new "user-sovereign, dynamic data" one for the Metaverse.
The Object Network implements that architecture and is an ideal foundation for building a single, seamless, open Metaverse - based around pervasive links that make the Linkiverse Challenge completely natural:
I build a nice garden. You do the same. We link our gardens together into a continuous landscape with a link in each direction, setting appropriate mutual permissions. I see that your garden has a tree in it whose leaves I like, so I walk across and grab the link to one of the leaves, then walk back to my own garden and swap all the leaves of my tree to this one. It's still your tree and the leaves can only be changed by you, so if you change the leaves to red in autumn, mine change too. A bird flies from your tree to mine.

Find out more

Pictures and slides on the Object Network Metaverse
Many more broader and deeper articles and presentations
Let me know what you think! There's a contact email at the base of the page.

Duncan Cragg, 2022. Contact me