First of all, thank you, Jérôme, for accepting to be part of the CEDAR Project's review meeting as an invited expert. And thanks also for caring to write down your comments on what was reported.
Most of what you wrote makes sense, from your persective, although there are some misunderstandings on your part concerning some of our work.
For one, the experiments we conducted, the results of which we presented in the Language Track aimed at evaluating the reasoning module used by the triple-stores, and not the triple-stores themselves. This is an important point. We used CEDAR as the reasoning module in Jena's triple-store manager. We also used the latter with its own reasoning module. We also used Jena's triple-store with Pellet as the reasoning module. The results we obtained were compared with those obtained by Jena's internal reasoner as well as Jena with Pellet. In the same manner, CEDAR can be used with other triple-stores that provide a plug-out/plug-in interface for independent reasoning modules.
Note that the reasoning module of a triple-store has no way to communicate to the triple-store any information concerning the latter's triple management. A triple-store management system does its own internal optimization of triple-oriented queries using its own SPARQL query processing system. Part of this processing is triple data indexing.
However, we wanted to test how much additional speedups in ABox querying one might yet gain if the reasoner could help the triple-store it is plugged into as an external reasoning module using bit-vector encoding for a code-driven indexing of the set of triples comprising the ABox to be queried. So we simulated a basic SPARQL query processor for simple conjunctive queries taking advantage of this code-driven indexing as explained in Sections 4.3 and 4.4 of CEDAR Technical Report 12.
Concerning your (JE's) comment:
Meanwhile, there remain questions to be precisely and definitely answered: what are the characteristics of the language and those of the solved problem on which it is so good? This would help putting such results in perspective and characterise the very contribution of this work. Moreover, there seems to be some room to extend it, both on the encoding side and on the language side, e.g., encoding disjointWith in ontologies. On the problem side, it would be useful to evaluate the gain beyond the Jena store and try with more optimised triple stores because there is a possibility that the gain provided by the CEDAR Reasoner is erased by optimisations of such stores.
The general answer is that what we have done is expose a technology for implementing basic taxonomic reasoning. We have also shown that it can accommodate roles as set-valued features to allow more expressive concepts such as existential and universal role-concepts used in several DLs. We demonstrated the sort of speedups this implementation technology can achieve.
What your are saying is a fair assessment. However, methinks, if I were one of the implementors of the systems we tested (or of other similar systems that we didn't), I would be quite curious to check out this method to see whether I could take advantage of it. Hence, the "very contribution of this work," that you are asking for is our implementation technique and our results for everyone interested to check out.
Be that as it may, to address this and your other interesting "suggestions for future plans," it would nice if we could pursue this work to give heed to what you suggest. Unfortunately however, we have no funding, nor any other project platform, that would allow us to carry on with many of the ideas we demonstrated during the CEDAR Project.
So this work is made public and documented for anyone to do with it whatever it's worth.
As for the remark concerning DisjointWith, you are overlooking the fact that it is simply not needed in a closed world as we (and many actual taxonomies and most ABox-oriented reasoning) do by default.
Thanks for the pointer to Chekol's thesis. This is very interesting work, and it looks like several of his ideas for processing the graph patterns of RDF-based queries are complementary to our concept-based encoding, as well as other work of mine (hak's).