I want to ask some question regarding the development of my push-to-talk application for Android. Recently, I've been developing push-to-talk Apps for Android. I use DatagramSocket to send voice and receive voice as a Packet over Local Wireless network (W-LAN). I use peer-to-peer network, so no server.
I don't have a problem with the code, but I don't understand the basic of VoIP theory, so I want to ask some question, hope somebody can give me simple answer :)
1. Is my push-to-talk Apps considered a VOIP-based?
2. There is several VOIP protocol such as SIP, H.323 and many more. If my PTT Apps considered a VOIP-based, and I use Socket-Packet (UDP, am I right?) to exchange voice, then what VOIP protocol that I use? Is it considered RTP protocol?
I would like to understand the theory behind my PTT apps, I understand my java code, but I don't have proper VOIP knowledge. I've tried to find some information in google, but I still don't understand what is the relation between my PTT Apps and the VOIP technology. Thanks before, I'm new here and sorry for my english!
I. "VoIP" is a very broad term, but, if your app transfers voice over IP network, it's definitely VoIP one, despite it may use totally proprietary protocols (as e.g. Skype does).
II. VoIP stack is basically split to two meta-layers - 1) signaling and 2) media transport. Each of them is in turn consisting multiple own layers (e.g. for SIP: session, dialog, transaction and transport layer). Examples of signaling protocols are H.323, SIP, MGCP. The most standard media transport is RTP. You can use your own transport; RTP applies specific restrictions (as AVP profile) but is compatible with variety of libraries and other implementations.
There are protocols which use the same socket and the same transport type for both signaling and media; the widely used one is IAX. Most others separate signaling and media, so sockets are separated and likely they have different type. A standard-compliant SIP implementation shall function over both UDP and TCP, and switch to TCP for large requests (>=1300 bytes by default); SCTP is also suggested. For all transports, protocol implementation details as retransmission policy and request timeout are specified in different way, but there is no principal problem with using any correct L4 protocol.
The totally another story is with media transport (under RTP or equivalent). Here, typical TCP manner to transfer all data at the cost of floating delays is really nasty for our ears. TCP is good for bulk traffic class, as file transfer or database interaction. In interactive communication between humans, we prefer more noise and sporadic voice loss than a voice strain. So, TCP is a very bad choice here, and a synchronous transport class shall be used, and UDP is the good default choice. SCTP can also be used as media transport, but with limited retransmit option (not all stacks support it). (There are attempts to use TCP to punch through NAT points but all this is an act of despair.)
If your application supposes sending a voice message more than to one recipient at a time (i.e. a kind of broadcast or multicast), this rejects use of connection-oriented media transports, effectively retaining only UDP. This also requires proper negotiation at signaling level.
III. Selection of voice codec is very platform-specific, I don't use which ones are native for Android. In "big" VoIP, there are licensed set and free set, with very small intersection (AFAIR, G.711 and GSM). Despite this, there are good codecs (e.g. Opus) which can be adapted into wide range of requirements, including partial packet loss.