FAT32 can support 2 terabytes. OK great. What does that cost the manufacturer of those drives? Nothing, because it has nothing to do with the actual hardware.
FAT32 isn't a "drive", it's a file system. Just like FAT16, NTFS, ext3, etc. They wrote a standard method to handle files on a system, and in this case they wrote in large overhead into their standard for future growth. Essentially a "software solution" of just giving the ability to read larger clusters.
You're right that it has nothing to do with hardware, either does the 1394b standard.
What does it cost to R&D and produce 1.6 mbps devices that won't even use 800 mpbs? A lot more than it's worth. When the devices come, the support will follow shortly after.
Your asking the cost of a device that can transfer data at 1.6mbps (surely you mean GB/s) "that won't even use 800mpbs?" I don't get the "won't use 800MB/s" thing. What do you mean by 'won't use'?
You also mention the devices should be made that are fast and then a standard should evolve around it? How you R&D a high speed device w/out the standard first is a cost I'd like to see, probably far more expensive to make your own standard instead of just licensing a tried and true already existing model.
In talking about the "cost" of making your technology fast enough to support future implimentations. It might've cost them a fortune to get that jump from 800MB/s to 1.6GB/s, and then ten times more for 3.2GB/s. But I doubt that. Given the "doubling nature" of the speeds I'd wager it's more likely just a tightening of the algorithm/encoding they use to transfer things, more of a "firmware adjustment" then a redesign of the physical hardware.
In fact, 1394b uses the same essentially wiring as 1394a. Two twisted pairs, but now uses a full dual simplex tranmission--meaning that data is constantly being sent over both pairs in either direction, like a highway. It's actual a simplification of the previous standard, which made it more efficient. Not a redesign or change in the actual hardware.
To help reach the GB/s level they've gotten a helping hand from IBM's 8B10B encoding scheme (another software level change), it's the same encoding that Gigabyte ethernet and Fibre channel uses.
The only thing you need to change about 1394b to go from 400MB/s > 800MB/s > 1.6MB/s > 3.2MB/s is the medium through which the data is transferred. All are still twisted pairs, all are still using the same protocol, they just have different plugs and less resistance as they go on (copper wire > plastic optic fibre > glass fiber).
Anyways, I pulled a lot of this from reading a presentation by a Microsoft Corp. IEEE 1394b guy and Steven Bard from Intel which was presented at WinHEC 2000. You can find it
here, there's a bunch more about BOSS, FOP/PHY, beta mode vs. strobe, etc.