[Cu-wireless] VoIP

[Stephane Alnet]

Hi,

Since I guess I was volunteered for heading this topic, here are some 
thoughts. ;D

* Documentation: voice (PSTN or VoIP) is a complex subject by itself. 
We need to better understand what "voice" entails; some of the topics I 
think we should document are:
   - VoIP signaling: architecture of H.323 and SIP(*);
   - call setup & call control, features (eg: what is call transfer, how 
does it work? what does a call setup look like in H.323 or SIP? what 
the heck *is* a call setup? :D )
   - VoIP bearers: RTP, codecs, events (DTMF, etc.);
   - quality of service (end-to-end): issues with voice (impact of 
delay, jitter, drop, ..), strategies (marking/queueing/etc.) -- most of 
this will not be wireless-specific.
   - fax, modem. Video. >:>
   - existing "open" projects in the works related to VoIP 
(openh323.org, gnugk.org, kphone, vovida.org, etc.).

This doesn't have to be big (we're not writing an encyclopedia), 
probably just a collection of pointers to other sites with more 
information for people to read, and a few people who read them and can 
talk about them to the rest of the group..

(*) IMHO we should try & justify our choice of protocols-set, if only 
for grant-writing purposes. For example, I don't think a call-agent 
based solution is in line with the goals of the project so that makes a 
case against MGCP & other call-agent based protocols. Etc. This could 
be an outcome of this work.

Also let's be careful about which pieces are available in source. 
(http://www.voip-info.org/wiki-Open+Source+VOIP+Software has some 
starting points, I guess.)


* Voice over Wireless network: I don't think we really need to spend 
too much time on "wireless phone" technology (PCS, CDMA, GSM, etc.); 
however it probably would be a good idea to get a "state of art" of 
issues related to (a) "voice" over 802.11b in an ad-hoc context, and 
(b) "voice" in a zeroconf context.

   I can think of things like:

   For (a) [VoIP/802.11b/ad-hoc]:
    - quality of service over a wireless shared media; IEEE 802.11e 
could apply, but we don't have APs: what does that mean? ...
    - path selection for voice RTP packets on a long-range wireless 
network (as far as I know this topic has never been covered): most 
802.11b research on QoS considers the endpoint access as the problem 
(and assumes a wired backbone); but we will have wireless as our 
*backbone*: what does that mean for route selection?
Are the path selection criteria for voice traffic (small packets / low 
bandwidth) the same as for data? [e.g.: 802.11's 1Mb/s may not be that 
good for data traffic but good for RTP voice bearer traffic if it's 
reliable..]
Do we need RSVP? And what would RSVP mean in a wireless, dynamic, 
ad-hoc environment?


   For (b) [VoIP/zeroconf]:
    - addressing / how does a device know the location of other devices? 
... locating a GK (H.323) or a proxy (SIP) in a zeroconf context; 
network partitioning and call routing; number assignment; directories.
    - PSTN gateway location/registration (think zeroconf), selection, ..

Probably more...


* Getting it to work: also based on the first point, we should try to 
get something up & running, even if it doesn't address all the problems 
we can find!

Possible subtopics:
    - hardware required for a PC-based endpoint;
    - potential grant / donations from HW vendors for trial purposes;
    - network components required for call routing (GK or SIP proxy);
    - PSTN gateway(s)...




I don't think this is even close to a beginning but we can try. :)  I'd 
like to propose that we start by the first step (documenting/ 
understanding) while keeping the end goal in mind.

We'll probably need a place to store the information -- wiki or 
something like that. What are the "official" resources? Then whoever 
has time to start researching could work on this..

S.


--
Mayotte - http://stephane.alnet.free.fr/