So what is it that I am ultimately proposing for semantic graph

Really, just that the 'lambda' operation be commutative:
@(a,b) -> @(b,a)

So from there we define a mobile, commutative nested order:

• Semantic layer
• BSON definition and expressions
• Ontology navigation, graph support
• B-Tree page definition

Not a difficult standard, since BSON is going to be a proper subset of existing byte code standards. Once traversal is commutative we get mobility. Let something like named graphs stew, be a competative value added. We know a sequence of B-Pages must map to a proper subgraph, so we have a simple mapping between the semantic object pointer and page pointers. That arithmetic makes it zippy fast to move sequences of pages around, the number of atomic data units per page to be decided. But, once we get a B tree definition, consider what a vendor like Cray and Yahoo combined can do, with the help of Cambridge Semantics.

How will this effect the market?
The difference between an ad and information goes away. Click thrus still generate ad revenue for authors. So business intelligence becomes a game of click thru fees. To the extent I can develop new and informative information, I can put it on the click thru pay list, before I blab.

For example, Cray would prefer I write proper white papers on semantic machines, and let them have first rights. So I just put some blog posts on the inside sale graph, Cray can buy for a week, or six weeks in advance all topics and software on semantic machine. Yahoo would manage their web bots to police the thing.

The atomic unit model is the subject,predicate,object triple, which at the graph layer is a triplet. Object is a nested bounding segment identifier computing in units, of fixed width, atomic triplets. Bound the basic unit to a 2**N address size. So the machine opens a subsegments, fills it with 32 byte BSON chunks, then closes it, the object is updated. The B-tree layer goes from object pointer to page pointer/triplet index.

Avoid decommuting the algebra. We can do a commutable keyword/predicate match, fill containers, use graphs as mobile objects, and rerun named graphs. When BSON injects asymmetry is when they do, all graph needs are a few counters.

Consider Cray's point of iew.
Having a subset of BSON, a B-Tree format, a JSON typeface and commutational traversal, then Cray has great options for formatting schema over their high parallel systems. They can run high performance, short sequence BSON expression in their muti-processors. The managing linux processor perform query optimization, then runs a B-Tree result straight into their hundreds of BSON machines with shared streams of disk based B-Tree streams. They real freedom to innovate, their company becomes more valuable. What they offer Yahoo clients is very special, a Watson like machine. Yahoo become the on line Bell Laboratories for corporate sponsors, high click thru fees.