[OT] USB KVM switches

From: Tony Duell <ard_at_p850ug1.demon.co.uk>
Date: Wed Feb 2 19:14:09 2005

> I like USB. Of course I don't develop for it :-)
> It's not supposed to be easy to develop for, but easy to use. It

But it's not easy to use -- if you want to do something the manfacturer
didn't intend (which for me is most of the time!)

> rarely screws up -- reporting some instance of a problem with it
> in this list is not enough to condemn it out of hand. It's silly.
> * It never requires adapters.

Only because adaptors for (say) linking 2 slave devices plain don't exist.

> * It provides (just adequate) device ID info for OSs to find the driver for.
> * Throughput is excellent (for most things).
> * It's bidirectional.
> * You never have to worry about connector sex.
> * Connectors are utterly standardized!

I'd much rather have to take a soldering iron to change an RS232
connector than do battle with USB...

> * It (largely) manages it's own address space.
> * It translarently supports multiple identical devices (if the
> driver does). Multiple pointing devices, multiple keyboards is routine.
> Try that with serial printers!

One of my PDP11s has 32 serial ports and that is nowhere near the
maximum... There is no good reason why all of them couldn't support printers.

> * It provides enough electrical power for many products, increasing
> with Moore's law.

Personally I'd rather give each device its own PSU...

> * There is only two variants of it and they are (in my experience) upward
> compatible (and maybe backwards).

At the moment

> * It really does handle everything from keyboards to hard disks with
> one architecture.

That's a minus point in my opinion. I've always found a 'one size fits
all' solution rarely does anything properly.

> * It is not made for "end users" to hack.
> * We are not end users.
> No one ever built commercial computer gear for the likes of us. It

Presumably that's why DEC gave away handbooks describing how to make
Unibus and Qbus devices. Why IBM sold a prototyping card for the PC (and
for the AT). Why HP sold an interfaceing manual for the 9100 and why they
produced an excellent 'Introduction to interfacing' for the 98x0 series
(giving schematics of the 2 main hackable HP interface modules for those
machines). Why HP sold the 82166C HPIL development kit (which, although it
contained 4 of the HPIL interface chips also included enough info to make
an HPIL device totally from scratch if you were mad enough). And so on.
Back then you were _expected_ to want to interface the machine to
something non-stnadard.

> was mainly historical confluences that made computer interface
> available to people w/o full laboratories. MOST computer
> interfaces are reasonably inscrutable; SMB, SNA, ethernet,
> whatever from whenever. Note I said reasonable; I know there are
> people here who could make an SMB interface from TTL but your
> criteria isn't the manufacturers. I see no one complaining about
> how complex ethernet is to make hardware for; you buy a card or a
> chip.

Or yuo do what 3 rivers did and make an ethernet interface from a 2910
sequencer, microcode PROMs, a few PALs (mostly for the CRC) and a lot of
TTL.... But I digress.

Point is, ethernet was rately, if ever, the only interface on a piece of
classic hardware. I'd haev no problem with USB if it didn't claim to be a
replacement for RS232 (it ism't, for a lot of reasons).

One last thing. In 20 years time I'll bet USB chips will be hard to find.
But I'll still be able to make RS232 compatible devices (heck, I've built
the level shifters from discrete components, I can bit-bang asynchronous
data in/out of any microcontroller port pin).

Received on Wed Feb 02 2005 - 19:14:09 GMT

This archive was generated by hypermail 2.3.0 : Fri Oct 10 2014 - 23:37:35 BST