Heh - Mike was exaggerating when he said unintelligible. While we are not masters of that paper, we worked through it and understood the algorithm. We could take a simple example and apply the steps - this is a very different understanding than someone who studies and focuses on this field, yes. Given time, we could have done the implementation without the python code - Mike's recollection of the early part of this story is heavily 2nd hand.
However, even understanding the algorithm (if not masters of all the concepts behind it), there was still a large gap to implementation. The solution that was used allowed us to focus on pieces of that problem and accelerate development fantastically.
Lucene has some of the best tests in open source software IMO. We are confident in this code - whether it takes 2 or 3 people to properly maintain or not.
The option before was a completely non scalable joke fuzzy query or nothing. Now you have this option. Great. A little magic? Sure. Great :)
However, even understanding the algorithm (if not masters of all the concepts behind it), there was still a large gap to implementation. The solution that was used allowed us to focus on pieces of that problem and accelerate development fantastically.
Lucene has some of the best tests in open source software IMO. We are confident in this code - whether it takes 2 or 3 people to properly maintain or not.
The option before was a completely non scalable joke fuzzy query or nothing. Now you have this option. Great. A little magic? Sure. Great :)