Re: ANNOUNCE: GNOME Speech API (GSAPI) - v0.6.0

[Michael Meeks <michael ximian com>]

Havoc,

On Fri, 2003-08-01 at 17:37, Havoc Pennington wrote:
> > > Why is this done in IDL? Are you expecting to use it out-of-process?

	Yes - precisely; there is an abstract speech server, that runs out of
process and does dirty things - like forking festival / linking to
chunks of mess etc. and then a far cleaner IDL interface to that.

> > Because the initial version of GNOME Speech (v0.2.x) use a client/server
> > model (Bonobo/ORBit2). I've tried to abstract v0.6.0 so that if/when an
> > alternative comes along, then it will require minimal changes to the API.
> 
> When you say "an alternative" you mean an alternative to what?

	An alternative to the existing gnome-speech interface.

> Basically the IDL interface is going to be too hard to use from C
> without a C library wrapper.

	That is how it is used currently; indeed there is only one minor issue
which is selecting the user's preferred sound backend - that can be done
with a standardised gconf key though I think.

>  The extra lib dependencies for all of
> Bonobo will also limit your audience and limit our future flexibility.

	I think you are confused. If the functionality is necessary -
code-reuse of something that implements that is the sanest / safest way
to go. Of course, one could re-implement a new application specific IPC
/ plugin / abstraction infrastructure each time one wanted something
like this - but ...

	Notably, 'all of Bonobo' is three smallish modules above the C library:
glib, ORBit2 and libbonobo, only glib/ORBit2 being really necessary for
client gnome-speech usage. Furthermore, CORBA allows trivial re-use by
Java, and the existing implementation of the FreeTTS backend in Java.

	Hrm,

		Michael.

-- 
 michael ximian com  <><, Pseudo Engineer, itinerant idiot