One should not underestimate the amount of secondary legacy software out there that rely's on the old OS
I don't I can think of 4 applications our engineers use that can't be ported onto modern laptops because they rely on the Rs232 port. And upgrading the software costs a 5 figure sum. And they need the internet to fire the data up to the OEM's database. Must go and clone them again actually. 2 of them are NT4 one is win95 and the other 98. Including software upgrades it would cost in the region of 50k to replace them.
There is a huge problem with this. The people who typically write firmware for embedded systems typically dont understand how high level operating systems work. They are used to writing software that talks directly to hardware as opposed to working with a layer of hardware abstraction. I know, I was the same once. So with win 95 and earlier, you could write software that talked directly with the UART for serial comms, but come Win NT you could no longer do this.
Enter the USB to Serial. First, you cant talk directly to the serial port. Second, there is another layer of abstraction that is the USB part, which essentially simulates a com port.
I have spent a lot of time, rewriting software that was written by hardware engineers so it works with modern OS's. Its not that hard. We have had 20 years now to sort this issue out. The problem being, most industrial hardware is designed to last more than 20 years. Any industrial hardware designed to last that long is not going to be cheap. AT some point, business just has to fork out the money to upgrade software. What annoys me, is most software from that era can be rewritten in a few weeks, it should not cost that much to replace.