[naasking]

This was illuminating:

> Universal Object Reference (UOR) provides a single format to publish, retrieve, and interact with any content. By “any content” we mean literally anything.

Then comes this:

> UOR has 4 principles that enable truly universal support for any and all content. Those are: > > 1. All content is formatted into a Directed Acyclic Graph (DAG)

So then UOR can't describe a cyclic graph? Doesn't sound like "anything" to me. It sounds like you might have to represent cyclic data using embedded runtime objects.

[tgv]

I suppose you have to describe back-references with an id to a node, or something like that. Anyway, you can describe all graphs in an object with two arrays.

[naasking]

Sure, but this violates the purpose of UOR, which is ostensibly to expose structure in a systematic way to make them understandable to automated systems. Encoding a cyclic graph makes the cycles opaque without adding this semantic information somehow. That's why I said maybe the embedded code that UOR permits might be the way to do it.

[vinceguidry]

The framework can be extended for new types of object.

[naasking]

But why build the core on a DAG when it's clear that cyclic references are everywhere?

[vinceguidry]

Because non-acyclic graphs aren't addressable.

[naasking]

I don't see why not, as long as you specify a preferred root and relative path from that root.

[vinceguidry]

With cycles in the graph, there's no single path to each node, no matter where the root is. Multiple paths to a node is the exact opposite of addressability. As an example, try to devise an algo for enumerating nodes and their paths from root.