In theory this sounds really cool.
However, without even trying it in practice I can already forsee one absolutely horrific problem arising: who takes all the notes and records the details so that when you come back to the (newly-established-tonight) village of Newton after a real-time year's play elsewhere (maybe only a month or two in character time, however) the road west still leads to Oldton, the inn still needs repairs to its roof (or is now undergoing such), and the village wisewoman is still named Florene and is in fact a hidden 5th-level Druid - etc., etc., etc. on both a micro and macro scale? Because I agree with [MENTION=22779]Hussar[/MENTION] (alert the media, it doesn't happen often!

) in that once something is established through play it is then locked in as part of the game world.
In a pre-built setting, whether it's the DM's own or a pre-canned one e.g. FR, a fair bit of the heavy lifting on the macro level (maps, town names, sometimes history and cultures, etc. etc.) is already done, meaning the note-taking can stop at the micro level and more focus on what the characters do in the world instead of also trying to build the world itself on the fly.
And I can't speak for you lot, but I know that when I'm running a game I usually have to do a lot of talking, and I can't talk and write at the same time. Thus, were I to run such a build-it-on-the-fly game, it would grind to a halt every couple of minutes while I made notes on what was just done and-or established. So much for any flow to the game...
I mean, look at the test example we've seen here with the king and drakes etc.; that'd represent about a page of notes I'd have to make on the fly
to cover what arose from a single encounter!
Lan-"but if you can afford to hire a recording secretary none of this matters"-efan