[asaph]

> Employing this interpretation, he found the Panini's "language machine" produced grammatically correct words with almost no exceptions.

If there are exceptions, I would think the problem isn’t solved. Perhaps the article is being loose with the wording. Close only counts in horseshoes and hand grenades.

[chmod775]

You're speaking a language that has some rules with almost no exceptions.

But almost means one can easily list the remaining special cases.

[asaph]

What’s the threshold of special cases allowed where the problem is considered solved?

[bigbillheck]

Buddy, I'm speaking English, it's full of exceptions.

[082349872349872]

Close also counts in computation, where we routinely handle infinite (or at least combinatorially effectively so) numbers of possible inputs with a few general rules and a finite ("almost no", to use a technical term) number of exceptions.

[asaph]

Perhaps my impression that this grammatical problem should be treated with logical rigor, where even a single counter-example invalidates the solution, isn’t the way this is being treated. The way it was presented in the article was like it was a kind of logic puzzle, for which an elegant solution has finally been found. Sounds like it’s more like a better rule of thumb was discovered that reduced the error rate, not to zero.

[kuroguro]

It's also possible some of the writings have grammatical errors.

[asaph]

Well how would you know that if the rules of grammar are not fully resolved?