In the said article, you'll find this statement:
The only difference is that you have to set appropriate state of RTS line before sending/receiving any data. Because RTS line is very often used in half-duplex converters to switch TX/RX mode, many programs can be used without any modifications.
Then the author describes, how RTS can be controlled through the Windows API.
In other words, he suggests to set it by the application software. This is in fact a longwinded and slow method. You should also consider, that you don't have exact control, when the data is actually sendout by the Windows OS. If you wait until you expect the data has been sent, it may be too late, because you have blocked the answer to your telegram. In my opinion, it's a poor-mans TX enable control, that shouldn't seriously considered. And it can't work through an ethernet tunnel.
Only the processor of the ethernet module respectively it's UART "knows" excatly when to set TX enable. Many processors have a hardware option to send this signal, otherwise it's usually generated from the TX interrupt handler, possibly using an auxilary timer.
If no suitable TX enable is available, you would generate it in hardware, as already mentioned. You can simply use a monoflop or a RC filter to extend the active low TxD signal by a certain amount of time, e.g. one bit time up to a frame duration. You'll see, that this method can't give an exact enable timing, it's either dropping during "1" bits or lasting too long, bringing up the risk of collisions. But's it's the best available method anyway.
P.S.: My personal preference is to extend TxD only by a small amount of time, just sufficient to charge the bus lines to the correct level with a "1" bit, and then leave it to the RS485 pull-up and pull-down resistor to keep the idle level until the next active "0" bit comes.