I've been listening to a few podcasts recently, which were obviously recorded using Skpye or another VoIP technology. Probably Skype, judging from the bad voice quality - my SIP phone sounds much better than that, but of course SIP is not always practical.

The problem is that it's usually the interviewer who records the conversation, so it's the interviewee's voice quality that suffers, although it is in most cases the one that we want to hear.

How about recording both side's voices locally, and mixing the two at post-production time? That's a bit more work, but aligning the two tracks on an initial pulse or beep should make it easy to get the timing right, and depending if the recordings can be reasonably isolated, not much more processing would be needed. Worst case, you'd have to adjust each track's volume according to who is speaking.

That wouldn't work for interviewing people who don't have a clue how to record their own voice, but geeks should manage ;-)