John, I use serial consoles on the servers and blades I manage, so I don't need speakup on them. In the case that I do need speech on them, I use a rescue disc with speakup. I need speakup on my desktop and I do prefer hardware speech over software speech. I also tend to do experimentation on my desktop or rather a test machine I have for this purpose, so I need access to kernel messages for when something goes wrong. This is why I need hardware speech on these machines. I guess only time will tell whether serial ports completely disappear from desktop machines. My money is on the side that they will. Afterall, it is already happening. What is the ratio of desktops that have a serial port to those that don't? My guess is around 10 to 1 or maybe even higher. And we're not even talking about laptops where it seems everyone is resigned to the fact that software speech is the only option. If speakup could support usb serial ports, than it doesn't have to be this way. One could use a hardware synth when it is docked and software speech when on the move if usb synths were supported. Perhaps rs232 serial ports will never completely disappear. There might always be a niche for them. You can still get rs232 serial printers that are used in manufacturing, so there might also be a niche for computers with serial ports. But we will have to pay a hefty premium for them. The point here is that we can take a chance and assume that there will always be computers with serial ports and can continue using speakup as it is now or we can take the lead and acknowledge the fact that serial ports are uncommon and that for speakup to be useful as a kernel based screen reader that it will need some way of accessing serial ports that are not built-in to the motherboard. I am not just talking about usb, as it does seem to me the best way to go, but there are also pci express rs232 serial cards that can be used as an alternative if speakup is able to access them. Since there are few desktops available with built-in serial ports and perhaps none in the not-too-distant future that have serial ports speakup can use, then speakup isn't very useful as a kernel based screen reader. I think we are at a cross roads where speakup development can focus on supporting hardware speech on modern computers without serial ports or it can focus on software speech where being a kernel based screen reader is of little advantage and it might be better for it to be in userland where it would be easier to integrate into distributions.
