// Copyright (c) 2003-2009 Nokia Corporation and/or its subsidiary(-ies).
// All rights reserved.
// This component and the accompanying materials are made available
// under the terms of the License "Eclipse Public License v1.0"
// which accompanies this distribution, and is available
// at the URL "http://www.eclipse.org/legal/epl-v10.html".
//
// Initial Contributors:
// Nokia Corporation - initial contribution.
//
// Contributors:
//
// Description:
// This is the extra documentation for the RTP API
// This is the API for the SymbianOS RTP stack.
// To use this API, start by creating a RRtpSession
// A session encapsulates all RTP traffic to/from a particular port, and
// its associated RTCP traffic. 
// RTP traffic from this port is modelled via the RRtpSendStream class.
// RTP traffic from a particular remote host is modelled via RRtpReceiveStream
// RRtpPacket and its subclasses RRtpSendPacket and RRtpReceivePacket are
// the encapsulations of the RTP packets.
// The API uses a R class 'cheshire cat' idiom. All the R classes in
// the API are handles onto resources, or you may prefer to think of
// them as wrappers round pointers. As such they have the following
// properties:
// - They are very lightweight and can be copied around by value.
// - 2 copies of the same handle are 'the same thing'; if you do something
// to one of them it will apply to the other.
// - They need to be 'opened' before they point to anactual resource,
// and must be closed when the resource is no longer needed. Note
// that in some cases, the resource is considered to be owned by
// another resource, so there is no need to close it. In these
// cases, there is no close method and this fact is clearly
// documented. If an R class is closed, any other copies of the
// same handle will become invalid as they are handles to
// nonexistent resources.
// The RTP stack relies heavily on an event model based on callback
// functions. This has the following advantages over alternative methods, 
// such as signaling events via TRequestStatus objects or using mixins:
// - It is more flexible for the client; we can distinguish a large number 
// of different events and allow the client to chose if they want to
// receive notifications about them in 1 function or many, and how they
// want to spread different notifications over their object structure.
// To give an example, a client might be interested in all SDES messages,
// only in some types of SDES messages or in no SDES messages, and they
// might want to handle these messages in different parts of the code or
// all in the same part. THe callback model can support all these cases.
// - Many other RTP APIs on other platforms use a similar concept, so
// creating a simialr API makes it easier to port code that originated 
// on top of these APIs.
// There are 4 major concepts in the event model:
// - An event manager. The session, the send stream and the receive
// streams each have their own event manager. Callbacks registered
// on 1 manager don't apply to other managers. The manager is not
// exposed in the API.
// - An event. (TRtpEvent class). An event is generated when anything
// happens that the client might want to know about. The event can
// be thought of as comprising 2 numbers, the event type taken from
// TRtpEventType and another parameter whose meaning is event-type
// specific. For instance, for all the 'Fail' events it is a
// SymbianOS error code. Additional information might be implicitly
// associated with the event, but needs to be fetched from another
// object, for instance when processing a ErtpPacketReceived event,
// the RRtpReceivePacket object must be obtained from the
// Rrt-ReceiveStream.  Events are associated with an object; for
// instance receive stream events are associated with a
// RRtpReceiveStream. There are functions on the event for
// obtaining that object.
// - A callback function. A callback function is a static function
// with a particular signature.  Static
// functions are used because callback models using member
// functions in C++ either involve bending the rules of the
// language or a lot of code bloat. If a function is registered for
// a particular event, it will be called when that event
// occurs. One callback function can be associated with more than 1
// callback registration. Callback functions take a pointer argument
// which was supplied as part of the registration; normally a
// static callback function in a CFoo class would take a pointer to
// a CFoo* and call a member function on the CFoo object that does
// the real work.
// - A callback registration. A callback registration is the thing
// that glues all this together. The event manager contains a
// number of callback registrations, each of which binds a function
// and pointer (normally an object) to a particular kind of
// event.  Registrations can be bound to all events on a stream
// (except 'Fail' events), via the ErtpAnyEvent code, or they can
// be further restricted by a parameter whose meaning is event-type
// specific. (For instance for the ErtpSDES event, the parameter is
// used to express interest in only a particular SDES field.) The
// same callback function and object can be registered for several
// different registrations, possibly even on different streams.
// So to use the event model, decide which events you care about, decide
// which objects care about those events, and register functions in all
// those objects to be told about all the events. 
// If an event leaves, an Fail event will be generated with the leave
// code.  Handlers for error events must not leave, as this would
// result in an infinite loop if allwed. It is reccomended that error
// events are handled by a separate callback to other events, to avoid
// confusion about whether leaves are allowed. (This is why callbacks
// registered for ErtpAnyEvent don't get called back for Fail events)
// One-shot event registrations are registrations that are only called
// back once. For instance, an application might want to display the
// NAME SDES parameter but might only need to be told what it is when
// the first one arrives. A one-shot resistration will only be called
// back for the first NAME received.
// To register events, use one of the RegisterEventCallbackL
// functions.  These are template functions that are available in
// global and member-function versions. The reason for this is
// limitations in the handling of template member functions in MSVC6;
// in many situations calls to the member functions won't compile and
// the global functions should be used instead.
// Note that there are always a pair of functions; one doesn't take
// the 'aParameter' parameter, and defaults it to
// KRtpNoParameter. This is more efficient as the parameter is often
// not used.
// The callback function takes a pointer which can be of any type.
// The RegisterEventCallbackL functions are all templated so that the
// pointer you supply must be of the same type as the function
// expects.
// As RTP/RTCP is removable from ROM using the SYMBIAN_EXCLUDE_RTP_RTCP
// in the build rom command a stub is used to prevent linking failures.
// The "#ifdef RTP_Removed" is used to reduce the overhead of having an extra set of cpp files to create the 
// stub. Using "#ifdef RTP_Removed" the MMP files create an extra set of stubbed 
// dlls from the one cpp file.  The iby file then exports the complete or stubbed dlls depending on the 
// SYMBIAN_EXCLUDE macro.
// The Open session functions leave with KErrNotSuported when the stub is used.  Functions
// which can not leave ASSERT in debug mode as should not have been called since the Open and similar 
// functions should have 'left'.
// 
//

/**
 @mainpage Symbian RTP API
 The RTP stack's event model is described in the section '@ref Events'
 @page Events The Event Model
 @page Stub of RTP API with #ifdef RTP_Removed
*/