Experimenting with Speech-dispatcher Virtualized and Remoted (Long)

[Veli-Pekka Tätilä]

Hi, Due to wanting to run Emacs with Speechd-el in a Windows programming 
environment, I've been playing around quite a bit with the various 
possibilities. This isn't terribly important as such, but something that's 
more of the nice-to-have variety:

1. A Qemu VM running Ubuntu Server: This was my first choice such that i can 
SSH into the box for decent Windows Braille and very high quality 
magnification with font interpolation. Espeak and SPeechDispatcher should 
take care of the smart Linux speech. It being QEmu based means I can run it 
as a limited user and technically from a USB stick, too. Two problems:

a. I'm not sure whether it is Qemu being extremely slow, the sound card 
virtualization (it is a PCI-based Ensoniq) or what but it is so sluggish as 
to be almost unusable when speechd-el and EMacs are on. IF I just use EMacs 
over SSH with no reader, it works decently. I could use the acceleration, 
VmWare, Virtual PC or CoLinux, but would lose all portability and require 
admin rights, then.

I think I've ruled out the possibility of it being Espeak. I wrote some 50 
lines of Perl for a UDP server that just speaks everything incoming 
asynchronously using the default SAPI prefs as set graphically in Windows. 
IF I use the generic Speech-dispatcher module to echo and  netcat stuff to 
it, it works, but the delay is still apparent, even if the server runs 
natively on the very same machine.

b. Ubuntu Server even with the base system, SSH the no X variant of Emacs, 
and no swap partition, is still almost 500 MB in size. Not very small, for 
an editor.

2. Win32-based Emacs with Speech-Dispatcher: someone suggested having the 
Linux laptop of mine as a speech dispatcher server. I tried that and it 
mysteriously failed. I'm going with the default port and have allowed remote 
access. I've given my LInux laptop a DNS name using the Windows hosts file, 
too, and set an environment variable to that name - I can ping that laptop, 
that's the machine name, ok. When I try to connect to LInux it says 
connection reffused in Speechd-el but does try contacting the correct 
machine name. IS there  a default Firewall in Ubuntu Hardy, perhaps?

I cannot really use the Linux laptop as the editing machine, as my Braille 
display and external display for magnification are used by the Windows 
machines almost exclusively. I use that laptop more like a test bed for 
Linux and for that purpose it works better than a VM or Wubi. BrlTTY does 
work, though.

3. Virtualizing only the speech dispatcher server: This could be very, very 
minimalist. All I need is speech dispatcher and eSpeak. Maybe not even that, 
if I use my SAPI-based simple server on the Windows client machine - then 
the VM wouldn't need sound, even. I'd have to make sure that eSpeak works 
remotely to begin with. I bet the Qemu VM speed would not be the issue since 
the server does have to run neither Emacs nor Speechd-el. What would be a 
good distro for this kind of special purpose? Obviously I'd like something 
whose base install is as small as you can get yet, being a newbie, something 
you  can use apt-get in. Both Debian and Ubuntu are quite large, actually, 
even with the default setup.
and I havne't yet found an easy programmatic way of stripping them to their 
bare essentials.

Finally, I've looked into smaller Emacs clones only having the text editor 
essentials. But they lose Speechd-el, dynamic abbreviation and 
minor-out-line mode, with a custom regexp, which are the top 3 reasons why I 
want to use Emacs.

Well, hope this has been an interesting read and that someone can offer a 
littel commentry.

-- 
With kind regards Veli-Pekka T?til?
Accessibility, Apps and Coding plus Synths and Music:
http://vtatila.kapsi.fi