dportas:
After reading about the NoSQL alternatives in the referenced blog, I find it to be just another case of apples vs oranges on this forum.
I will say categorically that until vendors are available with NoSQL solutions such as are described in the referenced blog site, that most of the commercial sites and ALL of the military sites I know of would say No to NoSQL, mostly because of the "No Throat to Choke" principle. The risk factor for a mission-critical data management system is just too high. If you are going to put all of your business eggs in one basket, that basket had (pardon my language) DAMNED well better have a viable vendor name on it.
I understand that structuring data to fit a set-theory model is sometimes not easy. I understand that folks will seek solutions that have the effect of reducing their overhead, and data conversion is often classed as overhead. But based on risk-reward theory, some risks are never worth any reward if the ultimate cost of the risk is total inoperability of your business after a failure or a determination of intractible data.
Further, associative-search algorithms laden with all sorts of heuristics and learning networks are great when they work. They suck when they don't, and because of that, a true general-purpose data management system doesn't exist. Many frameworks exist, but they require tailoring. We need sentient computers in order to reach the point of true general-purpose data management. While I don't say that can't happen, I'm fairly sure it hasn't happened yet. Without it, we have a gap between what any database program can do and what people want it to do.
I am not against ingenuity as a way to make things better, dportas, but in this forum you will find folks who don't always agree with the idea that doing away with relational models and SQL is viable for most businesses. Particularly since data mining (what you seem to be discussing) and business process modeling (what WE frequently discuss here) don't overlap that often. Different viewpoints.
Nevertheless, I'll agree that as computers become cheaper and clustering technology becomes more wide-spread, massive data management becomes more relevant in a global economy. Solutions are needed.
I don't think this next little problem has been adequately addressed in the solution you describe: Somebody's got to learn how to run those silly little data managers. The single-set models are simpler to manage than amorphous, overlapping multi-set models. And all too often, we see inexperienced programmers thrust into the position of becoming a DB manager or developer without having taken all the required course work. The harder you make that first step, the more likely you are to make a business fail. If know that, it's a step I would never take for my business. And that is a very big step for you to overcome in your movement to NoSQL systems.
Since I am not against progress, I will even wish you luck in getting where you are going, but I don't think we are quite ready for your solution. After all, being on the leading, bleeding edge of technology can be a painful experience.