Georg Zetzsche

Jul 22, 2016 11:38
@Raphael $P$ from the context-free is decidable: P(g,w,d) then says "d encodes a derivation tree for w in which a nonterminal repeats". For a suitable encoding, that is decidable.
Jul 20, 2016 21:32
Do you have an example of a pumping lemma that does not meet the conditions in my answer? Or do you know one where this is not clear?
Jul 20, 2016 21:32
If your point is that not everything that fits this definition of a pumping lemma is in fact a pumping lemma: That might very well be. The notion is is purposefully made extremely weak.
Jul 20, 2016 21:30
Think of the $d$ as a certificate that one can derive infinitely many words (a run with repeated states; a derivation tree with a branch repeating a letter etc.). And the decidability of P means one can verify that this is a certificate.
Jul 20, 2016 21:28
Take the NFA example I added in the text--Only sufficiently long $w$ admit a $d$.
Jul 20, 2016 21:28
@Raphael The $d$ serves as a witness for infinity. Not every $w$ has a $d$ such that $P(g,w,d)$:
Jul 20, 2016 20:33
@Raphael This notion captures the classical pumping lemmas. I made the NFA example explicit; does it make sense now?
Jul 20, 2016 20:33
@Raphael The definition tries to be as weak as possible. It just requires the existence of an infinite subset of $L(g)$ (you can call that set $L_{\text{pump}}$ if you like). It does not say anything about how these words are constructed. Note that this makes the notion of "pumping lemma" weaker and hence the non-existence stronger.
Jul 20, 2016 20:33
@Raphael I want it to be g to make the notion as weak as possible. Note that this merely requires that our language is infinite. Maybe there was a misunderstanding, I made the answer a bit clearer.