Brent Simmons, my former programming partner at UserLand, has been thinking out loud about rebooting Frontier. I've read his comments, let them sink in, and just emailed him a response. It makes sense to keep it all public, so here's a lightly-edited version of the email.
I get that you want it running on the Mac. I like the Mac, but I don't want to bet exclusively on it. I like JavaScript as a platform for that reason. (And of course JS code can run beautifully on the Mac, I think.)
I think the db-level code in Frontier is worth re-using. I didn't invent the allocation algorithm, it was in Knuth's book on algorithms. It's really stood up well over time. All it really needs imho is to be ported to JavaScript, and have the address size increased from 16 bits to 64 or even 128. That wouldn't add a lot of overhead, and it would leave the 2GB limit in the dust.
From there, I'd suggest factoring the outliner from Frontier, and polishing it up, really make it shine as a cross-platform thing.
That's where I started with Frontier in 1988. First the database, then the outliner and then I mixed and remixed. Those are the two core components.