BULL seems to work this way:

• Stuff needs to be said.
• There's not two equivalent ways to say the same thing,
  • though different sayings might occasionally point in the same direction.

But what needs to be said? Way too much, if you ask me.

Let me start by saying: “BULL is not really a language. It's a notation.” What I mean by this is that whereas human languages have very close written relatives, there's hardly a relationship between what is written and the information that is being transferred. BULL is quite the opposite; it's "just" a notation for the information that is being described. Nothing more. But neither does it lack anything in this sense. It's complete because you can describe "anything", human language being one of the more trivial but complex things you can describe.

Imagine the following sentence for example:

“There's something that I need to tell you: I love strawberries.”

There's a ridiculous amount of information compressed into this small sentence:

• A subject (I) loves strawberries. The sentence doesn't describe what or who “I” is. Neither does is describe what a strawberry is or what it means “to love” something. This knowledge is assumed from the reader.
• “I need to tell you something”: there's a difficult nuance in this phrase. Is this some type of confession? Perhaps it's an exclamation in gratitude upon receiving some strawberries. Either way, the answer depends on the context of the sentence. We also need the context to define who the speaker is.

We need context for almost anything. We need to know who is saying what. We need to know what language they're using. We need to know the story.

To encode all this information in a manner that is comprehensible to a computer is actually quite boring and painful, but it doesn't need to be. Enter the World Wide Web, a distributed collection of so-called “web pages” containing select phrases linking to other pages on the web.

The next version of the Web is called the Semantic Web (Web 3.0), meaning that it'll encode information in such a way that it's comprehensible by computers. This is a pretty big deal and BULL is based on this concept, except without all the creepy formalism and layer upon layer of needless legacy crap. RDF sucks for precisely the reason that URLs suck, for precisely the reason that URIs still suck and why Git is taking over from Subversion.

Decentralization is the key to the succes of the web, but DNS isn't decentralized. It's merely “distributed”. The only “cute” thing about URLs (don't get me started about URIs) is that name part, which doesn't do justice to the nature of the 'Net. The other cute thing that contributed to the succes of the web was the original simplicity of HTML. Everyone could learn it in minutes. HTML is no longer cute. It's encumbered by layer upon layer of crap and cruft like CSS and JavaScript. Then, there's RDF, which makes the thing that is most crucial for the web (linking) even more difficult than it ever was before the web was invented.

My proposal is to get rid of all that crap. While we're at it, let's get rid of all our programming languages and data formats as well. They suck. Besides, they aren't getting us anywhere. We don't need languages. We already have enough of those, most of them even before we ever dreamed of computers. We just need a notation for a universal data model. A temporary notation, until computers have gotten fed enough information to start acting intelligent of their own accord.

XML is not a universal notation. It doesn't come even close. Data isn't hierarchical in nature and never has been. To pretend otherwise is about as useful as following the Pope's advice on birth control: you still get screwed, just on more different levels at once.

RDF is close to the perfect data model, except that is uses identifiers which are masturbatingly formal, which is an indication that using URIs instead of URLs didn't keep anyone from actually identifying with locators. RDF still looks like it's describing locations, which is too bad because the universe doesn't consist of locations. It consists of relations.

So rid yourself of RDF's mental baggage and start afresh. All you need is this: a relation between point A and B. Let's call point A the subject and point B the relative. This is intuitive and at least doesn't sound like it is mostly meant to keep the terminology restricted to an exclusive and select group of computer scientists.

There we have our model. Let's expand it a bit by saying that there aren't really any subjects or relatives. Not only are they not locations or identifiers. They simply don't exists. Strangely enough, this doesn't mean that they can't be located or identified. Of course they can! That's what we have relationships for!

All we need to be able to do is to define and to traverse relationships. The rest is syntactic sugar for our notation, but our model needs to be able to comprehend this difference.

This already allows us to describe anything! There's nothing that can't be described in term of its relationship to other “things”. What it doesn't give us yet is comprehension. How does the computer know what everything (or anything) means? For that we just need to define a concise bunch of axioms, a set that's compact enough for us to easily describe to the computer. Then we can relate anything, literally anything we want to these axioms to make the computer understand our information. How's that for artificial intelligence?