It seems that a B&W G3 would be the best candidate, as it already has ADB. The only problem is that it doesn't have a floppy controller (or does it?), so that's out.
I'm sure that *technically* the B&W G3 has a floppy controller, since it actually used the same chipset (with some tweaks) as the Beige G3 (and the tray-loading iMac) but unlike the iMac the pins for it aren't brought out to a header connection anywhere.
It's sort of interesting/aggravating that Apple kept building Classic Mac serial ports into the chipsets of machines all the way up to the last OS 9 compatible G4 towers but bullheadedly only exposed them on the modem header. Of all the things that were dropped from the iMac compared to its predecessors I actually think the Localtalk port was the bridge too far, not the floppy. There were still quite a few LocalTalk networks deployed in educational environments in 1998, not to mention all the printers, MIDI devices, etc. When you consider a LocalTalk serial port was just a tiny little round DIN connector it almost seems petty they axed it. (And it *really* seems pointless when you think about how they deleted it from the B&W *but left the ADB ports*. I'm guessing the latter was to appease people with expensive specialized ADB input devices like tablets, but, well, a MIDI adapter was expensive too.)
one thing i was talking about with scott tonight is, maybe the reason the HD20 did not get any traction is simply because of Rodime.
No, forget about Rodime. Two things you need to remember about that Rodime hard drive:
#1: There is nothing about it that's intrinsic to how the HD20 works, IE, the "HD20 Protocol" parts of the HD20 are contained on the controller board with the IWM chip on it, there's nothing about it that's reliant on the Rodime hard drive except for the fact that the unit that shipped happened to use its oddball interface. (IE, it could have just have easily been built with a SCSI, MFM/RLL S506/whatever interface on the hard drive side.) And why did Apple use a Rodime drive with a bizarre proprietary interface?
#2: Apple paid Rodime to take one of their regular drives and slap a board with that proprietary interface onto it. The HD20's controller board was designed around the electrical specifications for the "Widget" drive Apple used in some models of the Lisa. They had a 20mb version of the Widget on the table, called the "Nisha", as far as prototype form which almost certainly would have ended up as the HD20's drive when Apple decided to drive a stake through the heart of their in-house disk storage unit in early 1985. (This is the same division that produced such winners at the Twiggy floppy.) With the Nisha suddenly no longer an option the panicked designers needed to salvage the HD20 project as quickly as possible, and in the course of that someone concluded it would be quicker to ask Rodime to adapt the Nisha control board to one of their existing drive chassis than it would be to redesign the HD20 to use a generic drive.
If there had been successors to the HD20 they almost certainly would have used an off the shelf hard drive interface instead of the Nisha protocol. But the reason there weren't any successors to it is once you admit a Macintosh *really does need* a hard drive as standard equipment you'll find it's far cheaper to just build a standard interface into it that bother with the ridiculous interface kludge that the HD20 is. The SCSI controller in the Mac Plus is pretty much just one chip, while an HD20-style hard drive requires an IWM, a microcontroller, and a wad of state logic to decode the floppy interface *and then* you need to have the hard drive controller hanging off the other side. The only way it'd make any sense to use it *internally* would be if you compressed the whole thing down to fit directly on a drive controller board, IE, you started churning out hard drives that *directly* spoke "HD20" (which the Rodime drive *does not*) with a 20 pin connector on the back. Apple would have to pay hard drive manufacturers to churn these things out (since obviously no one else would be using them), and when you consider the *maximum* speed potential of the interface is actually inferior to a brain-dead ST506 drive
by about a factor of ten it clearly starts becoming absurd. (The 10MB ST-412 drive in an original IBM XT transfers data faster than the HD20 despite being formatted with a 6:1 interleave ratio. A standard IBM AT controller with 3:1 is about four times faster and controllers for 1:1 were available. Add another 50% faster for RLL.)
Yes, the HD20 protocol (and SmartPort) works surprisingly well today with BMoW's Floppy Emu, but that's because with today's technology:
1: It takes only a little MCU and a CPLD to implement it, and
2: Using Flash memory makes the access latency/seek times very quick compared to an 80's vintage mechanical hard drive, which does a lot to make up for the slow data transfer rate.
Neither of these things were true in 1986. SCSI was a far better *and cheaper* solution unless you had no other choice. (IE, like in a 512k Macintosh or an Apple IIc with no slots.)
