Vertical vs square database technology and markets

That is a good division, finding the balance between flexible vertical formats and efficient square formats.  Database theory is all about how to complete a visit of every node on a graph using short cuts. When a group of nodes are nearly square, and we can impose constraint on their ranges, then we get SQL.  If you start with vertical format, you get MangoDB.  What the client and the micro processor need to agree on is the balance between squareness and verticalness. Like the file bar on our word processors, that drop down menu structure is a better quant of agreement.  It is a small finite graph.  That structure, that finite graph likely corresponds to something biological, the recursion limit to a set of counters in our brain, most likely.

So the market is going to come to a conclusion about a standard format that can embrace both the vertical and horizontal, a graph layer.  That standard is likely to be a super fast version of the java machine, with opcodes designed for hopping around graph structures. A specialized java code system just for the graph layer, designed to0 replace the compiled byte codes in sqlite3 and run as micro sequences at the match layer.  The semantic web will be built around a standard node. That node is likely to have two fixed length values and a string pointer, a triplet,  but when compaction and regularity are needed, square formats can kick in, and variations on the triplet. That is, the graph layer inherently has viability for squared data, especially a robust row pointer capability.

In large semantic ware houses, the risk machines will reserve a good chunk of their immediate cache and pointer space for this new code.  Most of these machines will never really paint a browser window, they will be Linux, tweaked, optimized for very fast execution of  node traversals.  There will be compilers that work very well with the triplet/nested format, generating TE components for large scale operations over the semantic web, web bots.  The underlying database bots will be very good at selecting the optimum schema an data grouping.