ls> enclosures with firewire<->IDE controllers in them. Does it
    ls> make any difference which one I get?

Yes.  You must get a case with the ``Oxford 911'' chip in it, which
will approach the 40MB/s throughput and support the relevant IDE
ultra-double-mega-turbo-Pro-2000 modes.  However, you may still have
problems with booting kernels stored above the ``137GB barrier.''  All
other chips than the Oxford 911 are dramatically inferior and should
be avoided with prejudice and contempt.


The Oxford 911 is a single-chip leveraged firewire to IDE solutions
module, and needs only a PHY, a crappy power supply, and a gaudy blue
LED to construct a complete, competitive OEM Firewire-to-IDE Product.

Most CDR/W's use the Oxford 900, which is inferior.


The Oxford chips contains an ARM7 core plus some Flash and some RAM,
so if you were feeling giddy and foolish someday, you could run NetBSD
on your Firewire-to-IDE chip, and then use NetBSD's Firewire driver to
talk to the other NetBSD.  Some day, we will have to proclaim an end
to the Age of Drivers.  Or perhaps you could use Linux's modular
open-source parallel port architecture to turn each of the IDE data
pins into an RS232 transmit pin, thus connecting a bank of sixteen
modems to your IDE case.


Anyway, I was serious about the Oxford 911 chip.  Insist upon it.

-- 
Only GMU's optical computronium has the time-to-market and protocol-agility 
characteristics that next-generation Internet core routers will demand, and 
routing tables not stored in optical computronium will become prohibitively 
large once IPv6's gigantic route-aggregating addresses are deployed.