[Opensim-dev] Voice Module

[Teravus Ovares]

Did I say TCP?  XD
-Teravus


On 2/27/08, Teravus Ovares <teravus at gmail.com> wrote:
>
> NAT traversal, what a fun topic.
>
> See http://en.wikipedia.org/wiki/UDP_hole_punching
> See http://www.bford.info/pub/net/p2pnat/
>
> The second of the two links goes into real detail.
>
> In TCP it's mostly useful when 'both' parties are behind a NAT router,
> however.   Less useful when just one is because the one computer that's
> behind a NAT router can be told to do the 'connecting' to the public one.
>
> The problem, is, it's a hack :D.  Sure, Skype use this hack to peer2peer
> connect callers.
>
> Are we planning on having direct end2end communications without a
> publically available audio proxying server?
>
> Remember if we are, we'll still need the publically available NAT
> Traversal coordinator that does the 'trick'.
>
> Best Regards
>
> Ter
>
>
> On 2/27/08, Dr Scofield <DrScofield at xyzzyxyzzy.net> wrote:
> >
> > Sean Dague wrote:
> > > I've spent a little bit of time looking at the Voice Module that just
> > > got moved into trunk from the RealXtend branch, and I think it's
> > headed
> > > in the wrong direction.
> > >
> > > Embedding voice in the simulator seems like the wrong direction, and
> > > much less modular than we were heading for.  Plus, I think that under
> > > any reasonable load, it's going to go useless *real fast*.  Given that
> > > chat messages can often take 30 seconds in a loaded message, nothing
> > > like voice really belongs there.
> > >
> > > I think a much better approach would be to carve off voice
> > functionality
> > > into a seperate VoiceServer, which would let it run on a dedicated
> > > machine, ensuring low latency.  I also think that the SLVoice approach
> > > of sending all the streams to the client and letting the client do the
> > > mixing is a better load approach.
> > >
> > ansi and i are working on getting the voice caps working from the
> > opensim side, so that we can then substitute the vivox daemon without
> > the SL client noticing (that's the plan anyhow).
> > > Perhaps a hybrid model where after significant avatar moves avatar
> > > possitions were sent to the Voice Server, so that you don't have the
> > > issue where your ears can be fully devoid of your camera, would be
> > > best.  But at the end of the day doing mixing on the server just seems
> > > wrong.  Even skype offloads this to clients to do.
> > >
> > completely agree.
> >
> >    cheers,
> >    dr scofield/aka dirk
> >
> > --
> > dr dirk husemann ---- math & computer science ---- ibm zurich research
> > lab
> > RL: hud at zurich.ibm.com - +41 44 724 8573 -
> > http://www.zurich.ibm.com/~hud/
> > SL: drscofield at xyzzyxyzzy.net ---------------------
> > http://xyzzyxyzzy.net/
> >
> > _______________________________________________
> > Opensim-dev mailing list
> > Opensim-dev at lists.berlios.de
> > https://lists.berlios.de/mailman/listinfo/opensim-dev
> >
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://opensimulator.org/pipermail/opensim-dev/attachments/20080227/f32228d5/attachment-0001.html>