Single-wire step and direction

I'm just kicking around an idea here I'd like to pass by everyone.
What would you think about coming up with a new step/direction
interface standard for motor drives?

The idea would be to combine step and direction information on a
single signal line. The advantages would be:

1) More reliable: No DIRECTION input to miss-wire, lose connection on
or damage with wrong voltages. No STEP to DIRECTION setup and hold
timing requirements, no STEP to DIRECTION crosstalk issues. A single
twisted-pair connection per drive would suffice.

2) Cheaper: One less input connector terminal per axis. A single
channel opto-isolator instead of a dual-channel one. One less
buffered, filtered and clamped input circuit to protect the digital
logic from the outside world.

3) Easier to use: There are only 2 possible wiring permutations with
2 wires; signal and common. There are 6 possible permutations with
STEP, DIRECTION and COMMON. Only one is correct in each case; 1 out
of 2 in the former, 1 out of 6 in the latter.

4) From the CNC program perspective only 1 output bit would be
required per axis instead of 2. CNC programs have to have a timer to
set the pulse width anyway. Changing that timing depending on
direction should be trivial. The freed-up interface bits (4 bits for
4 axies) could be used for other purposes.

The concept is simple. A step pulse narrower than 2 uS is decoded as
a CW step. A step pulse wider than 2 uS is decoded as a CCW step.

This is very easy to decode. The leading edge of the signal would
trigger a 2 uS one-shot. The signal would also go to the internal DIR
input of the drive. The one-shot timeout (Q not) would go to the
internal STEP input of the drive. If the signal pulse was less than 2
uS, the DIR would have gone 'high' before the step pulse active edge.
Otherwise it would still be 'low' on the active step edge.

This would require a single resistor and capacitor (RC=2uS) and a
schmidt trigger IC like a 40106, 74HC14 or similar. Maybe $0.15 in
parts minus the stuff no longer needed.

This is being put out as just an idea. Standards are notoriously hard
to change or introduce because of the "chicken or egg" effect. No one
will build drives for which there are no CNC programs to run them; no
one will write a CNC program interface for which there are no drives
to use it with.

Mariss