[Speex-dev] Library split (poll)

Thorvald Natvig speex at natvig.com
Thu Jun 29 07:58:06 PDT 2006

> Hi everyone,
> In the 1.1.x branch, I've kept adding more stuff to libspeex:
> preprocessor, AEC, etc. I'm now considering moving all those to a
> separate library (libvoip, libspeech, whatever). Anyone on this list has
> good reasons I should consider for either splitting or not splitting
> libspeex?

I'd much prefer it if it was still just one library. At the moment, I 
bundle the library with my Win32 applications (and one file is better than 
3), and for Linux compiles, it's much much easier to check the version of 
speex and then do a few appropriate #ifdef around the code, than having to 
check the version of each of the libraries (at compile time) and have 
different code for each possible combination of library versions that you 
wish to support.

For space-constrained projects, you can use a static library and a 
linker which only pulls out the object files actually needed (or 
alternately compile a version of the library for that specific purpose).

Actually, I'd kinda like to see even tighter integration between the 
different parts. At the moment, there's FFT back and forth at the end of 
echo / preprocess / start of encode.. Couldn't we just use the 
frequency-domain data from the preprocessor directly in the encoder?

More information about the Speex-dev mailing list