On Wed, May 01, 2013 at 05:46:28PM +0200, Samuel Thibault wrote: > John G. Heim, le Wed 01 May 2013 10:42:39 -0500, a ?crit : > > They did, however, take the time to criticize the speakup code itself. > > I'm afraid that's for a good reason: speakup works around the kernel, > that's not an approach that can work on the long term. Samuel is correct. I haven't heard anything on lkml indicating that there is a vendetta against speakup being in the kernel; it is just a very complex issue to solve, and no one has found a solution yet. Basically, staging/speakup/serialio.c needs to be completely rewritten, or possibly thrown out. The problem is that it uses hard coded addresses for the serial ports instead of communicating with them through the kernel serial port driver. That is why we are having more and more systems lately where speakup is having issues with serial ports. This is also why Samuel correctly says that speakup is working around the kernel. The proper solution to this issue would be to have a way that speakup can interface with the kernel serial port drivers. It will take some rewriting on the kernel side, as well as in speakup, to give us that functionality. It all comes down to time, and figuring out how to make this happen. William
