@HMC
Thanks for the excellent answer, that clears pretty much all my questions so far. I guess next step to understand better Tau is to get familiar with Idris and read Per Martin Lof's Type Theory (BTW for anyone else interested, there is a copy online
there. AFAICT this book is out of print, so you will likely not find it in the normal distribution circuit). That's going to keep me busy for a while. I'll come back with more questions once I have digested the underlying theory.
@Ohad
I've been playing a bit with CWM lately, and I'm underwhelmed by how the negation as failure under a closed world assumption is being handled. Unless I've missed something, it seems that the only way to express a negation using basic CWM builtin semantics is to parse a N3 document into a formula with log:semantics and test for the absence of a specific clause in that formula using log:notIncludes. That seems to be confirmed by a few sources: Norman Walsh's
blog and some
chat he had with Dan Connolly (one of the authors of CWM). These sources are old but CWM hasn't evolved much since and I haven't been able to find any other way by looking at the documentation and test files. If that's still the case, it means that in CWM negation as failure can only be achieved on static documents included in a file and cannot be done on the general context of the reasoner which is always considered to be open world. So much for the so called "Closed World" machine Huh...
Anyway, since your code examples so far have been compatible with CWM, I was wondering how you planned to handle the negation as failure in Tau from a syntactic perspective. If you already have some idea of what the syntax will be, can you please give an example? I'm thinking of a simple case: build a rdf:List of all objects that do NOT have a specific property set. Let's say for instance all objects that are not a fish to continue the example in your (btw very convincing, congrats!) LetsTalkBitcoin interview.
I'm starting to build an application that I will port to Tau as soon as it's functional, and I need a way to express a negation in a closed world. I could move up the food chain to something like OWL and use set semantics to get my way but I'm concerned that this is like going in the wrong direction on the completeness-vs-consistency spectrum and is going to make it more painful to backport the application to Tau in the future. Can you recommand some existing RDF-flavored language that is close to the target in Tau, and that allows to derive negation as failure in a closed world kb without having to rely on static files to do so?
Of the course the ideal would be to start directly with Tau if the reasoner is already functional. But it would also need builtins.. Maybe I can help you with that if you already have a clear idea of what builtins will be available. If you are planning to adopt a subset of the CWM builtins among other things, that could be a good start.
