Nil Geisweiller

Yeah, I suppose to be handled in general requires to tell whether an Evaluatable is functional (as opposed to non-functional, i.e. state-sensitive). For instance `IdenticalLink` is functional and its outgoings...

Till this is fixed, the obvious workaround is to use a dummy formula.

I agree, there are striking resemblances between OpenPsi and the URE. The URE control mechanism is actually a specialized form of OpenPsi. OpenPsi rules however are declarative and probabilistic, while...

The difference is that in the first case the URE will consider no premises, and in the second case it will consider n premises.

What would probably be needed as well is to have atomese reduct works for type links, so for instance ``` TypeSetLink TypeNode "ConceptNode" TypeNode "ConceptNode" ``` could be reduced to...

You could use a separate atomspace instead, but if you're using it then it's better to let it there for now.

Looking into https://en.wikipedia.org/wiki/Incremental_computing could provide some help.

The backward chainer already somewhat supports that, i.e. the user can pass an auxiliary atomspace where the trace is stored. This is for instance used in the inference control meta-learning...

No inference control meta-learning experiments have been attempted with the forward chainer yet, so trace support is lacking.

The following constitute a start of an API to extract FC traces https://github.com/ngeiswei/reasoning-bio-as-xp/blob/6006874a9918b29510080d5ef43ffc5b7fc809e4/bio-as-utils.scm#L348-L479 It just needs some work and being moved to the URE.