Interesting news: it seems that finally a commercial developer has attempted to go beyond SQL and create a true relational database implementation. (see discussion at http://www.dbdebunk.com/x_trdbms_impl.htm).
The company is called Alphora (www.alphora.com) and the product in question is called Dataphor. To my knowledge, this is the first serious commercial attempt at providing a true relational method of handling data. The interesting part about this is that this company is not a traditional RDBMS vendor at all. They actually approached this from the perspective of needing a better application development environment. They discovered some of C.J. Date's writing, including "The Third Manifesto", and decided this was the best possible method. Their description:
Complete applications virtually “fall out” of the database definition, yet they can also be tailored to fit the requirements of any project.
This is exactly what I have been thinking would be possible from a complete relational data implementation. Dataphor actually modeled their query language on Tutorial D, from "The Third Manifesto". This also includes object-relational type inheritance.
One caveat to this development. Dataphor is only a data manipulation environment, and lacks its own physical storage layer, at this point. Thus for data persistence, it uses existing SQL engines in the background. Again, let me stress that this is on the physical implementation level, not on the logical data level. As Date, Darwen, Pascal, et all... say, a database can do anything it wants on the physical storage level, as long as it provides relational constraints on the interactive level. The only problem is that it still suffers from some of the inefficiency of current SQL implementations, for physical storage. However, during the data's "lifetime", it can actually reside in live memory, so performance can be good, up to the moment you need physical storage. Also, the physical storage is an open API, so you can write your own.
This sounds like a very interesting project. The only disappointment on my part is that there is currently no project by any open source developers that attempts anything equivalent to this. It seems to me that it would be a natural fit for PostgreSQL, but the developers are apparently more interested in "good SQL implementation" as being good enough for the present.
Also, Alphora only runs on Windows, being a .NET add-on. I hope they realize that there is a significant market for we Unix lovers too. If it could provide me a full relational query environment, with an C/C++ API, and support for PostgreSQL on the backend, I would willingly pay for it. (And you know the PHP and Perl developers would have a module for it in no time)
On a side note, there is a "placeholder" project at SourceForge, for anyone willing to assist in starting an open source implementation of Tutorial D. (Yes, there is absolutely nothing there at present) Leandro, the project maintainer, is simply hoping to provide a focal point where these things can be discussed, and perhaps even planned. I hope anyone here who is interested will provide whatever support they can. (that means you too, André ;)).