The EIA-485 standard just defines a driver/receiver chip spec with little comment on how you apply it.

I rarely end stories myself - let them run where they need to. Since "True EIA-485" requires wiring stub lengths of 0 inches (ie: no circuit board routing even permitted), it only exists in labs and simulations anyway.

We just make converters; so we have to support both. But I know of lots of process control equipment which has 4-wire RS-485. The Master device talks on 1 pair to all slaves, and all the slaves share the second pair to respond. EIA-422 doesn't define a tri-state (or high-impedence) transmitter, so true EIA-422 cannot do this. Since the slaves never see the responses of other slaves the data comm interrupt load is GREATLY reduced with this design. This is very important for single CPU slave devices. Some 2-wire suppliers avoid this by disabling their comm ports for 100-150msec each time they see a master request for another slave, but this can seriously cut down on your RS-485 through-put if you have a fast master.

The 4-wire variety also seems to be more forgiving of bad wiring - I even had one customer buy converters to make their 2-wire a 4-wire system because they needed to run faster than their 2-wire installation situation permitted - too many data errors at higher speeds and they didn't want to rewire the site. Should have bought good cable? I've yet to see a large outdoor site were the RS-485 doesn't end up on the same type of twisted pair, multi-core wire used for analog loops. But I'm sure some people somewhere do it perfect. Somewhere.