Hey, On Tue, 2003-08-05 at 18:11, Ville Syrjälä wrote: > On Tue, Aug 05, 2003 at 05:39:12PM +0200, Ronald Bultje wrote: > > On Tue, 2003-08-05 at 16:57, Ville Syrjälä wrote: > > > If anyone wants to write code that has a chance of working with all > > > sourcards alsa-lib is a must. I see no point in making it optional. > > > > Why? > > Would you prefer that all application writers would have to worry about > how the hardware expects the data? There are just too many different ways > soundcards work to make this viable. The fact that an additional layer is needed to unify different hardware, doesn't that mean that ALSA (kernel API) missed the big deal of being a kernel API? The kernel API is supposed to handle all this. If it doesn't, it sucks and needs to be replaced. That's why v4l2 superseeded v4l1. v4l1 is good, but v4l2 is simply better. > And the user has a lot of control on how alsa-lib does things. Take > channel routing for example. There's no way each application would have a > way to configure stuff like that. But the application still doesn't do that, but a different application or configure file does. So I see no different in the kernel-API vs. userspace lib approach. > Also alsa-lib can do software mixing. It's nice, but it's bloat too. > > *Proper documentation* is a must. A wrapper lib is just an > > implementation of the required documentation that never existed for > > ALSA. I think you're turning things around here. > > If you really wanted to you could write your own library to talk to the > alsa kernel drivers... "Proper documentation is a must". The ALSA kernel API is badly documented. (Well, no wonder people use that lib thing). Ronald -- Ronald Bultje <rbultje@xxxxxxxxxxxxxxxxxxx>
