Design suggestion: The server just for synthesis

[Michael Pozhidaev]

Hello, Hynek!

> Actually it's implementation as a text protocol has also been
> defined already in a draft:
>
>     
> http://cvs.freebsoft.org/doc/tts-api-provider/tts-api-provider.html#Text-Protocol-TTS-API

OK, we can easily consider it.

> Yes, both the API and its implementations are still in draft stage,
> so improvements are very welcome.

Hynek, can you give me some resources on freebsoft.org where I can do
this work? First of all - git repository. It seems to me the best way is
to save drafts into text files and edit them with discussing. Then this
git repo can be used and for source code control too. When text
documents are ready we can publicate them on the web. OK? If using
freebsoft.org is not convenient for you I can easily register new
project on sourceforge.net . What about 'ttsbunch' as name for new
package?

> I don't actually think we are talking about some entirely new
> server. It's much more about refactoring the existing Speech
> Dispatcher technology into three pieces, re-using its knowledge,
> its design etc.

Yes, we must use already created code so much as it possible. Also :
I have some ideas, what do you think about following:
I would like to use the minimum configuration data for TTS server. For
example the set of available TTS the server instance is serving must be
defined only by list of so-files installed in some special directory. No
registering in configuration files is permitted. It allows activating
new TTS only by installing corresponding package. I say 'apt-get install
foobartts-plugin' and it is everything required to use it. 

-- 
Michael Pozhidaev. Tomsk, Russia. E-mail: msp at altlinux.ru
Russian info page: http://www.marigostra.ru/