I'm a PBX tech (Nortel), so I may be a little biased. I'm also interested in the possibility of building and selling Asterisk boxes (as key system alternatives) because, with the proper cards, Asterisk supports traditional telephony and, since it includes voice mail, CDR, a conference bridge, etc, it could be very cost effective alternative.
What I'm not particularly interested in is using VoIP for businesses. Why? Because it does not consistently work, especially in small markets. In a small business enviroment there is too much out of my control -- and reliability suffers. And reliability is the major concern of business phone users, not whiz bang, gosh golly, gee whiz features.
The main problem that I see with Asterisk (and to some extent the whole Open Source community) is the constant tinkering mindset. "Open and flexible" is nice, but open source folks tend to constantly rebuild. Business users simply want certain features and they want the phone to work when they pick up the handset. Recompiling the Linux kernel five times a week isn't going to hack it in the telephone business. (Neither will the lack of good documentation.)
Asterisk, especially certain flavors like Xorcom, may be close for the key system, but as a PBX replacement they aren't even on the radar. Modern Nortel hybrid switches, like traditional PBXs, are highly reliable. Redundancy and hardware stability are the basic principles behind every PBX switch Nortel sells. New Nortel switches have added inter switch redundancy. The goal is to retain the reliability of the traditional PBXs on the hybrid IP platform.
I recently downloaded and read "Asterisk -- The Future of Telephony." I was very enthusiastic, and I still think it may be in my future, but the tendency of those who work with VoIP phones to pooh pooh reliability in traditional telephony, shows they are out of touch with the needs of businesses. Disconnect a business PBX from its power source for five minutes some time. You will probably lose your job -- or you'd better have a damned good reason. On the IT side, it seems quite acceptable to announce the network going down for installation of patches in the middle of the afternoon. That's part of the IT mindset -- it will not fly in telephony. Period.
VoIP promoters also want to pooh pooh the very real problems with QoS. Even the writers of the Asterisk book acknowledged that it was a real concern, but then ignore it. They also explained that IP packets are not condusive to real time speech. So, until there is a solution to these problems, I think any prediction that the PBX is dead is a bit exaggerated.
I think another problem for Asterisk (and other VoIP switch makers) is the very real reliability problems in the Cisco Avvid systems. Many corporations have "gone Cisco" only to return to a traditional PBX when the system did not perform up to past standards. In other words VoIP is gaining a very negative reputation. I think I've read recently that Nortel and Avaya IP switches are both now outselling Cisco's VoIP switches. There's a reason for that.
And one more point. The author suggests that the big PBX makers go fully VoIP. They tried going that direction (admittedly with the Windows platform) and it didn't work. Now that they've taken the NEC route (hybrid) their sales are growing. They've been able to retain reliability while supplying some IP features -- including IP phones.
Asterisk claims that its TDM support is "trasitional." Yet, to me, that's the attraction. I know the telphone network supplies reliable voice circuits, while the IP network is still struggling with QoS and trying to fit the square peg IP packet concept into the round hole of real time. Until there is a viable solution, in small markets as well as in big ones, VoIP will never replace the circuit network.
In short, traditional PBXs work, VoIP only sometimes "kind of" works. And, until VoIP foks really understand that reliability is the main feature required in a phone system, they will not take the market.