James Tauber wrote:
> Is it simply that the implementation complexity outweighs the perceived
> usefulness?
Yes. I do not think that this feature is required to declare a victory.
I also think that delay is more fatal than lack of features.
> Concur was as simple for me to implement as any other pattern (and
> considerably easier than overall non-determinism). I realise this was
> because my implementation is backtracking.
I strongly believe that each feature should be introduced after considering
how to create tree automata. Backtracking implementations are easier, but
they ocasionally perform exhaustive search and issue inappropriate
error message.
Cheers,
Makoto
