A common problem when trying to run several languages on the same virtual machine is to be able to interact between languages. In a perfect world, we would like this to be seamless and transparent. Let's take .NET as an example; you can call a C# class from your VB.Net program without any problem. This is possible because they both share the same type system which is the one specified in .NET, so it's often said that whatever language you can run on .NET, it will be C# since you'll have to match this type system. That's quite true.
The Array Problem
Neko is trying to reach language interoperability by data sharing. One common problem you have when interacting between languages is about arrays. Every language has arrays, but with different APIs. Some languages can resize the array, some can't. Some languages can modify the array, some can't. Some languages access arrays with an Object Oriented API, some don't... So how can you pass one array from a language to another?
One possibility is to have a super array that has all of these possibilities and is shared between all languages. That doesn't scale very well since you might have same method name with different behaviors depending on the language.
One other possibility is to be able to convert between arrays, but that doesn't scale well either since you need to add more conversion functions every time you're supporting a new language.
The Neko way of doing this is to provide a common data representation of the structure that will be shared between several language specific APIs. Because not all APIs are Object Oriented, a Neko array is not an object. Because not all arrays are resizable, a Neko array is not resizable.
Several languages then can share the same Neko Array Reference and wrap the data structure with their own API. The only thing needed is then a generic way to wrap a Neko Array with the Language API, and to retrieve the Neko Array from any Language-specific Array.
The Class Problem
Neko doesn't provide a fixed class system. It doesn't provide a way to check if an object is of the given class or implements an interface. It's up to the language generator designer to choose how they want their class system to be represented within Neko.
In some languages, function calls are type checked, and some not. By providing a runtime dynamic type system, Neko makes it easier for static and dynamically typed languages to interact together.
As a result, it opens up more flexible ways of encoding classes and doing type checking. It's not particular to any language, so you can roll your own, optimized for your own language way of dealing with classes. And if you don't have classes, even if you don't have objects, Neko is still suitable since it doesn't enforce any specific way of doing.
However, this might cause some problems in the way objects are represented at runtime between languages. The preferred way is to use Neko objects so you can seamlessly interact with other Neko languages.