Lewinski,
I've used several active x-overs and I divide them into 2 groups. The analog models included a tubed Marchand, an NHT x-2, and the internal active x-over in my Theta Casablanca. The Marchand was too noisy for me, but that might be unique to that piece, rather than the model, since I've been told that others had more success. The NHT is an excellent piece (and IMHO a great value), but I ended up pairing it with a Velodyne SMS-1. The resulting performance was excellent, but the set-up was a giant PITA (thanks to the Velodyne) and the multi box/cable arrangement was inelegant. The Theta is a really flexible and good sounding unit, too - and a one box solution.
I moved to an Onkyo 5509 pre-pro after I heard Audyssey xt-32. A lot of A/B testing at a couple of local places led to a loan of an Integra pre-pro which I felt was (much, to my taste) superior to the Theta. The crossover in the Onkyo (like most modern pre-pros) is effected in the digital domain. I ended up buying the Onkyo version because I couldn't hear a difference in sound between it and the Integra version, it was significantly cheaper, and I preferred its appearance.
To the x-over question: It's hard to separate the performance of the digital x-over (vs the analog x-overs) because they were integral (pardon the Onkyo Integra pun) parts of two very different pre-amps/pre-pros. In the end, I like the Audyssey powered one-box solution of a pre-pro for my main 2 channel system and that means a digital x-over comes as part of the deal. Whether it's the digital cross, the room correction, the balance of the preamp circuitry, or any combination of the above, the bottom line is that I found it to be a better mousetrap. Since I had already effectively stopped listening to LPs in favor of a server (and consequently didn't really have to deal with the cognitive dissonance presented by ADC and DAC for my analog LPs) I retired my ARC LS-25 and Joule LA-150 without much indecision. As always, YMMV.
Hope that helps. BTW, I've never heard the Dirac correction package, but I've done a bunch of research on it: There's definitely some interesting software there. If you do pull the trigger on Dirac, I'd be grateful if you posted your take on its performance.