I'm not educated enough with the production of "commercial software" to leave a a solid enough counter-argument to stop these people. They claim that a legacy client would require continuing support, which as far as I know is completely false. I consider old programs I've developed for old OSes that are still posted online but no longer supported by me to be "legacy", they work "as-is", and I've always assumed it to be as such.
This is true for standalone software to a certain extent (it could be possible that legacy software will not work anymore when other parts of the system are updated, such as the OS, drivers, hardware etc).
The problem with Steam is that it is an online platform, so there are two sides here:
- The Steam client software
- The server-side Steam services
As soon as you want to log in to Steam, you need to contact a service. When you want to download/install a game, you need to contact a service etc.
So the problem here is that if they want to keep supporting legacy clients, they have to keep the services stable as well... at least to the point that a 'legacy' Steam service exists alongside a 'cutting-edge' one.
I think this is the biggest problem in the software world today: nearly everything relies on web services.
A lot of software can't even be downloaded anymore. You get a simple installer 'shim', which downloads and installs the actual software when you run it. Of course this will break at some point in the future.