[Ekiga-devel-list] Pulse support and Networking (was Re: Reorganizing things)

[Dave Allan <dallan redhat com>]

I hesitate a little to speak here, as I'm not completely versed in the 
issues or the code, but here's a short statement of my experience with 
these two problems on Fedora 10 using the SVN head as of about a week 
ago.  The head from this morning crashes when I try to make calls, so I 
can't say if the behavior has changed since then.  See the rest of my 
comments inline, below.  I would be happy to file bugs against these 
behaviors if that would be helpful, or work with someone familiar with 
the code to troubleshoot them further.

yannick wrote:
> Peter Robinson a écrit :
> > > Sorry, it is a bad timefor me as my work is starting again; I'll have
> > > fewer spare time for 2 or 3 weeks...
> > >
> > > > (Except Peter: thanks for the nice words and help offer!)
> > > >
> > > > Le lundi 31 août 2009 à 19:58 +0200, Damien Sandras a écrit :
> > > > > Dear all,
> > > > >
> > > > > It appears clearly that since a few months I can dedicate less time than
> > > > > before to Ekiga. During the first 5 to 6 years of the project, I was
> > > > > dedicating all evenings to the project and nearly all week-ends, fully.
> > > > >
> > > > > It is now impossible for me to continue developing at that pace. If we
> > > > > organize things in a clever way, I think it will not be a problem. My
> > > > > intensive work probably hide a few organization problems.
> > > > >
> > > > > We now have a few very high quality contributors: they help the project
> > > > > move forward will less "devotion" from myself.
> > > > >
> > > > > The purpose of this e-mail is to identify the areas where people can
> > > > > help and how we should work. But before identifying those areas, I would
> > > > > stress on the fact that my wish is to release often, but with only a
> > > > > limited set of new features, that are well-tested. We have seen in the
> > > > > past that we had worked on many new features, half-finished, and that it
> > > > > was hard to stabilize them before doing a release. I don't want this to
> > > > > happen.
> > > I like this idea too, 3.0 was a pain to release. Beside, network (TCP
> > > support) and the audio part of Ekiga (e.g. pulse audio) needs some new
> > > features quickly.
> > >From a Fedora point of view the two "bugs" I see on a regular basis
> > are "Audio Problems" (will be fixed mostly by a Pulse Audio plugin set
> > as the default) and Network issues which are mostly Firewall and NAT
> > based. I think support for upnp would possibly fix most of those
> > issues for the average user. The would be the most useful features
> > from my point of view and the later would probably also help the issue
> > with ekiga.net "not working" mentioned elsewhere.
>
> I see pulse audio as a good thing for the Linux audio stack; it provide
> a unified test case of the major features and my hope is to see it
> improve the situation at the end. (and IMHO it really needs improvements...)

Pulse support in the SVN head as of a week or so ago is only partially 
functional.  When I place a call, I hear ekiga generate the ring sound 
until the other end picks up the call.  The other end hears only digital 
artifacts (soft blip-blip-blip noises) until I open the pulse audio 
volume control application, at which point the other end hears correct 
audio.  Closing the volume control application reverts to digital 
artifacts.  Reopening it causes audio to work again.  I never hear audio 
from the called party.  These behaviors are 100% reproducible.

> The network issue is also a complex matter. I agree upnp will improve
> the situation too. Still it will not be enough for some people. The
> software we compete with here (skype) is full of workarounds for
> hardware/network config *hostile* to VoIP. As people with software which
> do not work are always louder than people where it just work, the
> situation will still harm ekiga reputation. Beside, the security topic
> here is quite important and very complex, and this is something skype
> deals with obscurity. Quite frankly, I doubt if the skype protocol was
> open, it will be considered as secure... I've started to work on the
> topic, i.e. how to deal with hardware/setup hostile to VoIP. I've hope
> to publish a white paper in the coming months (I was full of hope to
> publish this for the end of august, but I was then busy with user
> support, bug triaging, packages, etc.). I do not see a good fix for that
> in the short term. Any help is welcome.

On Fedora 10, I have a machine with several interfaces including a VPN 
interface.  I am only intermittently able to register and make calls 
through my SIP server.  Checking the box in preferences at:

Edit->Preferences->General->General Settings->Network Settings->Enable 
network detection

seems to have a positive effect, in other words, ekiga seems more likely 
to register and place calls when that box is checked, but it's not 100% 
effective.

If I create a virtual machine on this physical machine, ekiga registers 
and makes calls 100% of the time.  Both machines are running F10, fully 
updated.  The significant difference is that the VM has only one 
interface, plus loopback, whereas the physical host has 5 interfaces 
plus loopback.

Twinkle registers and places calls reliably from this machine.  I don't 
know the twinkle code, either, but since it is FOSS, perhaps it would be 
useful to look at its code to see how it operates in the multi-interface 
case.

Dave