This file is a work list and specifications for the tasks which need
to be completed to get the git version of speakup ready for
prime-time.  Hopefully folks will undertake some of the tasks here and
help by submitting patches to implement them.  If you can think of
better ways of implementing any of these please speakup!


The install script will need to be made smarter so that the correct
speakup version can be checked out for the correct linux kernel
version.  It also needs some way to actually create a reject file when
it fails so it's easier to hunt down the patching trouble.


Various methods have been discussed on how to best rewrite portions of
speakup to implement the tty discipline for serial synths.  Currently
we have kludged the io port method we've always used by supplying the
parametres which have been removed from kernels starting at 2.6.22.
This however is only a temporary fix except for synths which use
non-standard rs232c ports.  Samuel has suggested using filp_open() for
module built speakup and possibly a virtual file system for the
built-in case.


Need to re examen synth probing to attempt to identify the synth by
the probing sequence it responds to.  It is at least important to
identify that you have the synth you have requested to prevent getting
the wrong serial port.


Speakup needs to be made into a kernel thread so kernel preemption and
other non-interruptable things won't drag the kernel down and
continuously break up speech output.


Speakup needs internationalization support.  It's current strings used
for notification need to be pulled together into an array of strings
which can be referenced by index thus providing substitutions from
other languages.  Each language's array of message strings could then
be loaded by /sysfs psudo files or by loading modules written for that
purpose.


A method needs to be developed to once again provide a way for speakup
to speak the kernel messages from boot up.  This might be done
possibly by reading the printk() circular buffer and sending it to the
synth.  It might also be done by once again capturing the early kernel
output and storing it in the speakup synth buffer until a synth is
registered.
