[pulseaudio-discuss] Pulseaudio and hal deprecation/DeviceKit.

[Lennart Poettering]

On Mon, 11.05.09 16:47, Maxim Levitsky (maximlevitsky at gmail.com) wrote:

> 1 - like he says, hal will still be there, just like libbonbo is.

I figure you mean libbonobo? The phasing out is mostly done AFAIK,
with the exception of the panel.

> 2 - there will be huge wave of very unpleasant bugs, and again there
> will be release or two containing most of them (remember folks, you
> don't have the money to buy every system/device) in the world to test.
> And while users do report bugs, its time consuming for user and
> developer  to do remote debugging.

HAL is pretty buggy too. It's sometimes ridiculous. For example, it
patches ACLs *after* it sent out signals that a device was created, to
the effect that when PA tries to open the device then it often enough
cannot because the access fails.

> Just an example, we have a linux lab at my university.
> Users bring usb sticks, plug it in, and it doesn't work.
> (I figured out that this is due to stale .hal-mtab file in /media)

Yes, HAL is broken in many ways.

HAL was a first attempt to figure out what we actually want on
Linux. Now we know that pretty well, so the second try is hopefully
going to be a bit smoother.

Also note that the process to swich to udev has been going on since
quite some time. In F11 quite a few subsystems have already been
switched over. A few are still remaining.

> Thus ONLY incremental changes are possible (and no 'buts' or 'ifs')
> Even if you create an experimental branch (like kernel mode-setting)
> it just won't progress, because developers are already full of bug
> fixing the current stuff that user use.  Even now, I hear that
> modesetting still crashes the system there and there.

Those changes are mostly incremental. HAL and the new udev/dkit stuff
do not conflict. Things are moved over bit-by-bit.

In the end in Free Software it's always at the developer's discretion
how things are handled. If you want to have a greater influence on how
things are handled, become a developer of the specific project too!
And I'll promise you you'll start to see things more like the other
devs by doing so.

> So back to pulseaudio, I think that what developers should do is to
> rewrite alsa mixer, and device enumeration.

Oh yes. But we as you noticed we are bit limited on workforce. And
that's even more true for Linux audio than most other things on Linux.

Lennart

-- 
Lennart Poettering                        Red Hat, Inc.
lennart [at] poettering [dot] net
http://0pointer.net/lennart/           GnuPG 0x1A015CC4