And Is it fast?

One of the open areas is the issue of speed. SQL databases have matured over 25 years now, so the expectations are high. While I think that the jury will be out for some years, here a couple of thoughts.

SQL databases are generic. They are not aware what data structure they are carrying, this is defined declaratively by the database schema. An SQL-based solution like TMRQL will have to fine-tune the generic engine by adding indices and/or caching techniques.

What makes me sceptical that this works is that access and query patterns for semantic web applications are not along tables for which most of the optimization mechanics of relational databases has been developed. They are along semantic axes (whatever that is), so not only along relationships such as instance-of and subclass-superclass but also application-specific ones. I seriously wonder how a relational database will react on a high number of self-joins there.

We also expect that TMQL queries can be sped up considerably if the TMQL processor knows more about the structure of the content. That is then maybe encoded in a TM constraint language or may be even hard-coded in a TMQL processor. It remains unclear, yet, how this can be converted into background knowledge to be usable for TMRQL.