Performance – As you point out, calling methods on interfaces is slow. I also showed this in my article on function performance. I assume you mean some kind of getter for the node data here. My approach requires no function call, getter or otherwise, but instead just accesses an untyped variable (*). This is consistent with the Array API in that it simply holds anything you want. Even if you did have a getter for the data, it too would need to be untyped so as to support any type and be portable and reusable, as you point out.

Number of objects – An interface doesn’t get around the creation of objects. You have to have an object at each node so that you can hold the next and previous pointers. The only way around this would be to require that the only data allowed to be stored in the list would have these pointers built into it. This is possible using an interface, but gets really restrictive when you consider that you have to create node wrapper classes a lot (eg. class StringNode, class IntNode, class UintNode, …). Still, instances of these node wrappers are also objects. So I’m not sure how you can save on the number of objects. Could you show an example?

PS: I think I’ve found your article about AS3 linked lists. I’d like to read a follow-up to it showing the complete linked list class and some usage examples. Do you have one? I’d like to learn more about your strategy. :)

Is it a performance issue? Interfaces are inherently slower and accessors are slow as well, but I find the flexibility of the interface much more portable and reusable.

Another downside to a node object is object creation. This implementation would require creating your value about, creating your node ‘container’ and then nesting your data inside of the node – doubling the amount of objects you have.