Open uzytkownik opened 3 years ago
What exactly is your question or your proposal? :smile:
Related idea: Coq's constructive epsilon module.
@MatthewDaggitt
I guess question:
In theory I have no objections to it. The usual approach is to put the type definition in Axiom.X
where X
is the name of the axiom that you would like to add to the system. You can then take the axiom in as an assumption to a particular proof in the library and hence remain --safe
.
As for the name, I'm sure there's probably an official one somewhere which I'd advocate using. Otherwise I might be tempted to go with something like the super-catchy CountableDecidableExistential
? Another question is would we want the type to be over naturals or over any type that is Countable
? The latter definition doesn't yet exist in the library...
It's called Markov's Principle
If we can prove
¬ ((n : ℕ) → ¬ P n)
for some decidable propertyP
we can get∃ P
The question becomes if this function terminates:
This property is much weaker than double negation axiom: