software speech for speakup

Kerry Hoath kerry at gotss.eu.org
Sun May 13 23:57:50 EDT 2001


Ok so that was all 1 big long line! I won't be slicing that message
down into smaller sections in ed os it remains attached below.

The fundermental problem with software speech and speakup is this:
speakup gets control extremely early on in the boot process, just after the
console driver or at the same time. At this point; no sound is initialized,
no hard disks are known, ho usb is active, no file systems
are mounted and therefore sound and viavoice can't yet be loaded.
Via voice is a huge application, and putting it into the kernel isn't an
option since it would make the kernel image too large and anything in kernel
image is unswappable and consumes memory.
Not to mention we don't have the viavoice source so we
can't even integrate it if we wanted to.

It may be possible to do something like keep speakup asleep until viavoice
is running, and make some shim between speakup and viavoice but this
is not trivial. What you are trying to do is take information
from kernel space (console driver) handle it with speakup, hand it to userspace
to a program that mightn't even be running anymore, have that program
synthesize the speech and pass it back to kernel land sound drivers that we
presume you loaded. This is going to make the performance of the system like
cyphoning honey uphill. Even assuming the speakup mods were made by somebody
in the forseeable future, there are many more moving parts to make work.
What if one of the tasks such as viavoice dies or sound drivers unload, how
do we tell the kernel to tell user space to tell the kernel to tell user space
that something ahs gone wrong?
Emacspeak is a user application. It calls a speech server
to interact with its talking device and assumes that sound and viavoice are
in top shape. When the speech server crashes emacspeak respawns it.
What you are asking is for speakup to become re-enterant, the ability to put it
to sleep and wake it up at will and the ability to talk to it from user space
despite the fact it is kernel code and have that kernel code talk back to
user space. This requires a complete redesign of speakup
and although it may be possible, so is Bill Gates giving away all his
money and becoming a hermit.
Even if Kirk changed his mind regarding viavoice tomorrow and coded flat out
until the project was complete it would require months of coding time before
the whole system was usable if indeed it could be done.
Remember Windows screen readers run in user land and although they hook into
the windows subsystems they are applications like any other. Speakup is in
the kernel itself and is part of the operating system.
Ever had jfw or windoweyes crash and lost your speech? Often you are left with
no clue as to why it happend and often jfw is unrestartable. If we were to
have this happen in Linux it may result in bits of the kernel becoming
unusable and could lead to an entire system crash.
My personal recommendation is to learn how to use emacspeak, preferably from
a seasoned emacs user and learn about term mode and shell mode.
W3 is a nice browser, vm works well and so does emacspeak. Once you have this
down pat, you can then use the c-mode in emacs to start writing the code
for speakup and take some of the weight off the existing coders <smile>
If it takes you a week to get emacspeak working for you, it will tide you over
until tuxtalk is ready for prime time.
There are other userland screen readers and one supports software speech can't
remember what it is called. Use that until the massive
modifications are in speakup itself around 2010.
If we get more coders things might go faster, but until then; you might need
to use another solution for accessability if you have no serial ports.

Regards, Kerry.
On Sun, May 13, 2001 at 06:13:40PM -0500, Gregory Nowak wrote:
> Hi All,
> 
> Ok, here is my penny's worth on software speech for speakup. I certinly don't mean to flame or unconstructively critisize here, so please read on if you're interested. There are some of us that don't have serial ports on a PC, but do have a sound card supported in Linux (based on some earlier posts I've seen on this list, I know I'm not alone in this situation). As a result, I would personally like to entirely blow away the other OS on such a machine, and dedicate it to Linux (simply because I'm getting tired of using the other OS on it, and because all its hardware is Linux compatible). I know that Kirk mentioned that he was working on a software synth that would work with speakup in the far future, and that he wouldn't write anything for IBM viavoice, because it wasn't an opensource product. However, as I stated earlier, there are thoes of us that would like to be able to use software speech with speakup in the very soon future (now). Yes, I know that I could use emacspeak which supports viavoice. However, I've recently downloaded it and played with it for two days (even read  the howto). Given a choice of access though, I would much rather stick with speakup. Thus, not writing a driver for a product that is not opensource (and so far for me works without a hitch) is a serious limitation to access. Speakup certinly doesn't have to be distributted with viavoice (emacspeak isn't), but it would be nice to have the option of using it. It shouldn't be that hard to modify a dectalk or doubletalk PC driver to work with the speech engine. I've taken c++ my junior and senior years of high scghool (AP computer science). Even so, I have somewhat of an understanding on how the sample programs work that come with the engine. I also plan to read the API docs, and hopefully learn more. As you can see, I'm not a candidate to write the speakup driver for viavoice, so I'm not volenteering. I certinly wouldn't mind switching to the opensource engine when it became usable. However ... ok, I've wined enough. Kirk, I guess I'm simply asking you if you could possibly consider changing your mind regarding viavoice as a speech engine for speakup for now at least. If there is anyone else here that agrees with me, please write so that we could see how many more takers there are, and maybe try to persaude Kirk some more to change his mind. Thanks for reading.
> Greg
> 
> 
> 
> _______________________________________________
> Speakup mailing list
> Speakup at braille.uwo.ca
> http://speech.braille.uwo.ca/mailman/listinfo/speakup
> 

-- 
--
Kerry Hoath: kerry at gotss.net
alternatives: kerry at gotss.eu.org or kerry at gotss.spice.net.au
ICQ UIN: 8226547





More information about the Speakup mailing list