Merge Call completion support into trunk.

From Reviewboard:
CCSS stands for Call Completion Supplementary Services. An admittedly out-of-date
overview of the architecture can be found in the file doc/CCSS_architecture.pdf
in the CCSS branch. Off the top of my head, the big differences between what is
implemented and what is in the document are as follows:

1. We did not end up modifying the Hangup application at all.
2. The document states that a single call completion monitor may be used across
   multiple calls to the same device. This proved to not be such a good idea
   when implementing protocol-specific monitors, and so we ended up using one
   monitor per-device per-call.
3. There are some configuration options which were conceived after the document
   was written. These are documented in the ccss.conf.sample that is on this
   review request.
		      
For some basic understanding of terminology used throughout this code, see the
ccss.tex document that is on this review.

This implements CCBS and CCNR in several flavors.

First up is a "generic" implementation, which can work over any channel technology
provided that the channel technology can accurately report device state. Call
completion is requested using the dialplan application CallCompletionRequest and can
be canceled using CallCompletionCancel. Device state subscriptions are used in order
to monitor the state of called parties.

Next, there is a SIP-specific implementation of call completion. This method uses the
methods outlined in draft-ietf-bliss-call-completion-06 to implement call completion
using SIP signaling. There are a few things to note here:

* The agent/monitor terminology used throughout Asterisk sometimes is the reverse of
  what is defined in the referenced draft.

* Implementation of the draft required support for SIP PUBLISH. I attempted to write
  this in a generic-enough fashion such that if someone were to want to write PUBLISH
  support for other event packages, such as dialog-state or presence, most of the effort
  would be in writing callbacks specific to the event package.

* A subportion of supporting PUBLISH reception was that we had to implement a PIDF
  parser. The PIDF support added is a bit minimal. I first wrote a validation
  routine to ensure that the PIDF document is formatted properly. The rest of the
  PIDF reading is done in-line in the call-completion-specific PUBLISH-handling
  code. In other words, while there is PIDF support here, it is not in any state
  where it could easily be applied to other event packages as is.

Finally, there are a variety of ISDN-related call completion protocols supported. These
were written by Richard Mudgett, and as such I can't really say much about their
implementation. There are notes in the CHANGES file that indicate the ISDN protocols
over which call completion is supported.

Review: https://reviewboard.asterisk.org/r/523


git-svn-id: http://svn.digium.com/svn/asterisk/trunk@256528 f38db490-d61c-443f-a65b-d21fe96a405b

2 files changed, 417 insertions, 0 deletions

diff --git a/doc/tex/asterisk.tex b/doc/tex/asterisk.tex
index 8097d14ed..0427389b1 100644
--- a/doc/tex/asterisk.tex
+++ b/doc/tex/asterisk.tex
@@ -147,6 +147,9 @@ reference purposes.
 \chapter{Security Framework}
   \input{security-events.tex}
 
+\chapter{Call Completion Supplementary Services}
+  \input{ccss.tex}
+
 \chapter{Development}
   \section{Backtrace}
   \input{backtrace.tex}
diff --git a/doc/tex/ccss.tex b/doc/tex/ccss.tex
new file mode 100644
index 000000000..cfe07cbe0
--- /dev/null
+++ b/doc/tex/ccss.tex
@@ -0,0 +1,414 @@
+\section{Introduction}
+
+	A new feature for Asterisk 1.8 is Call Completion Supplementary
+Services. This document aims to explain the system and how to use it.
+In addition, this document examines some potential troublesome points
+which administrators may come across during their deployment of the
+feature.
+
+\section{What is CCSS?}
+
+	Call Completion Supplementary Services (often abbreviated "CCSS" or
+simply "CC") allow for a caller to let Asterisk automatically alert him
+when a called party has become available, given that a previous call to
+that party failed for some reason. The two services offered are Call
+Completion on Busy Subscriber (CCBS) and Call Completion on No Response
+(CCNR).
+	To illustrate, let's say that Alice attempts to call Bob. Bob is
+currently on a phone call with Carol, though, so Alice hears a busy
+signal. In this situation, assuming that Asterisk has been configured
+to allow for such activity, Alice would be able to request CCBS. Once
+Bob has finished his phone call, Alice will be alerted. Alice can then
+attempt to call Bob again.
+
+\section{Glossary of Terms}
+
+	In this document, we will use some terms which may require
+clarification. Most of these terms are specific to Asterisk, and are by
+no means standard.
+
+\begin{itemize}
+\item CCBS: Call Completion on Busy Subscriber. When a call fails because the
+recipient's phone is busy, the caller will have the opportunity to
+request CCBS. When the recipient's phone is no longer busy, the caller
+will be alerted. The means by which the caller is alerted is dependent
+upon the type of agent used by the caller.
+
+\item CCNR: Call Completion on No Response. When a call fails because the
+recipient does not answer the phone, the caller will have the opportun-
+ity to request CCNR. When the recipient's phone becomes busy and then
+is no longer busy, the caller will be alerted. The means by which the
+caller is alerted is dependent upon the type of the agent used by the
+caller.
+
+\item Agent: The agent is the entity within Asterisk that communicates with
+and acts on behalf of the calling party.
+
+\item Monitor: The monitor is the entity within Asterisk that communicates
+with and monitors the status of the called party.
+
+\item Generic Agent: A generic agent is an agent that uses protocol-agnostic
+methods to communicate with the caller. Generic agents should only be
+used for phones, and never should be used for "trunks."
+
+\item Generic Monitor: A generic monitor is a monitor that uses protocol-
+agnostic methods to monitor the status of the called party. Like with
+generic agents, generic monitors should only be used for phones.
+
+\item Native Agent: The opposite of a generic agent. A native agent uses
+protocol-specific messages to communicate with the calling party.
+Native agents may be used for both phones and trunks, but it must be
+known ahead of time that the device with which Asterisk is communica-
+ting supports the necessary signaling.
+
+\item Native Monitor: The opposite of a generic monitor. A native monitor
+uses protocol-specific messages to subscribe to and receive notifica-
+tion of the status of the called party. Native monitors may be used
+for both phones and trunks, but it must be known ahead of time that
+the device with which Asterisk is communicating supports the
+necessary signaling.
+
+\item Offer: An offer of CC refers to the notification received by the caller
+that he may request CC.
+
+\item Request: When the caller decides that he would like to subscribe to CC,
+he will make a request for CC. Furthermore, the term may refer to any
+outstanding requests made by callers.
+
+\item Recall: When the caller attempts to call the recipient after being
+alerted that the recipient is available, this action is referred to
+as a "recall."
+\end{itemize}
+
+\section{The CC Process}
+
+\subsection{The Initial Call}
+
+	The only requirement for the use of CC is to configure an agent for
+the caller and a monitor for at least one recipient of the call.
+This is controlled using the cc\_agent\_policy for the caller and the
+cc\_monitor\_policy for the recipient. For more information about these
+configuration settings, see configs/samples/ccss.conf.sample. If the
+agent for the caller is set to something other than "never" and at
+least one recipient has his monitor set to something other than
+"never," then CC will be offered to the caller at the end of the
+call.
+
+	Once the initial call has been hung up, the configured
+cc\_offer\_timer for the caller will be started. If the caller wishes to
+request CC for the previous call, he must do so before the timer
+expires.
+
+\subsection{Requesting CC}
+
+	Requesting CC is done differently depending on the type of agent
+the caller is using.
+
+	With generic agents, the CallCompletionRequest application must be
+called in order to request CC. There are two different ways in which
+this may be called. It may either be called before the caller hangs up
+during the initial call, or the caller may hang up from the initial
+call and dial an extension which calls the CallCompletionRequest
+application. If the second method is used, then the caller will
+have until the cc\_offer\_timer expires to request CC.
+
+	With native agents, the method for requesting CC is dependent upon
+the technology being used, coupled with the make of equipment. It may
+be possible to request CC using a programmable key on a phone or by
+clicking a button on a console. If you are using equipment which can
+natively support CC but do not know the means by which to request it,
+then contact the equipment manufacturer for more information.
+
+\subsection{Cancelling CC}
+
+	CC may be canceled after it has been requested. The method by which
+this is accomplished differs based on the type of agent the calling
+party uses.
+
+	When using a generic agent, the dialplan application
+CallRequestCancel is used to cancel CC. When using a native monitor,
+the method by which CC is cancelled depends on the protocol used.
+Likely, this will be done using a button on a phone.
+
+	Keep in mind that if CC is cancelled, it cannot be un-cancelled.
+
+\subsection{Monitoring the Called Party}
+
+	Once the caller has requested CC, then Asterisk's job is to monitor
+the progress of the called parties. It is at this point that Asterisk
+allocates the necessary resources to monitor the called parties.
+
+	A generic monitor uses Asterisk's device state subsystem in order
+to determine when the called party has become available. For both CCBS
+and CCNR, Asterisk simply waits for the phone's state to change to
+a "not in use" state from a different state. Once this happens, then
+Asterisk will consider the called party to be available and will alert
+the caller.
+
+	A native monitor relies on the network to send a protocol-specific
+message when the called party has become available. When Asterisk
+receives such a message, it will consider the called party to be
+available and will alert the caller.
+
+	Note that since a single caller may dial multiple parties, a monitor
+is used for each called party. It is within reason that different called
+parties will use different types of monitors for the same CC request.
+
+\subsection{Alerting the Caller}
+
+	Once Asterisk has determined that the called party has become available
+the time comes for Asterisk to alert the caller that the called party has
+become available. The method by which this is done differs based on the
+type of agent in use.
+
+	If a generic agent is used, then Asterisk will originate a call to
+the calling party. Upon answering the call, if a callback macro has
+been configured, then that macro will be executed on the calling
+party's channel. After the macro has completed, an outbound call
+will be issued to the parties involved in the original call.
+
+	If a native agent is used, then Asterisk will send an appropriate
+notification message to the calling party to alert it that it may now
+attempt its recall. How this is presented to the caller is dependent
+upon the protocol and equipment that the caller is using. It is
+possible that the calling party's phone will ring and a recall will
+be triggered upon answering the phone, or it may be that the user
+has a specific button that he may press to initiate a recall.
+
+\subsection{If the Caller is unavailable}
+
+	When the called party has become available, it is possible that
+when Asterisk attempts to alert the calling party of the called party's
+availability, the calling party itself will have become unavailable.
+If this is the case, then Asterisk will suspend monitoring of the
+called party and will instead monitor the availability of the calling
+party. The monitoring procedure for the calling party is the same
+as is used in the section "Monitoring the Called Party." In other
+words, the method by which the calling party is monitored is dependent
+upon the type of agent used by the caller.
+
+	Once Asterisk has determined that the calling party has become
+available again, Asterisk will then move back to the process used
+in the section "Monitoring the Called Party."
+
+\subsection{The CC recall}
+
+	The calling party will make its recall to the same extension
+that was dialed. Asterisk will provide a channel variable,
+CC\_INTERFACES, to be used as an argument to the Dial application
+for CC recalls. It is strongly recommended that you use this
+channel variable during a CC recall. Listed are two reasons:
+
+\begin{itemize}
+\item The dialplan may be written in such a way that the dialed
+destintations are dynamically generated. With such a dialplan, it
+cannot be guaranteed that the same interfaces will be recalled.
+\item For calling destinations with native CC monitors, it may be
+necessary to dial a special string in order to notify the channel
+driver that the number being dialed is actually part of a CC recall.
+\end{itemize}
+
+	Note that even if your call gets routed through local channels,
+the CC\_INTERFACES variable will be populated with the appropriate
+values for that specific extension.
+	When the called parties are dialed, it is expected that a called
+party will answer, since Asterisk had previously determined that the
+party was available. However, it is possible that the called party
+may choose not to respond to the call, or he could have become busy
+again. In such a situation, the calling party must re-request CC if
+he wishes to still be alerted when the calling party has become
+available.
+
+\section{Miscellaneous Information and Tips}
+
+\begin{itemize}
+\item Be aware when using a generic agent that the max\_cc\_agents
+configuration parameter is ignored. The main driving reason for
+this is that the mechanism for cancelling CC when using a generic
+agent would become much more potentially confusing to execute. By
+limiting a calling party to having a single request, there is only
+ever a single request to be cancelled, making the process simple.
+
+\item Keep in mind that no matter what CC agent type is being used,
+a CC request can only be made for the latest call issued.
+
+\item If available timers are running on multiple called parties,
+it is possible that one of the timers may expire before the others
+do. If such a situation occurs, then the interface on which the
+timer expired will cease to be monitored. If, though, one of the
+other called parties becomes available before his available timer
+expires, the called party whose available timer had previously
+expired will still be included in the CC\_INTERFACES channel
+variable on the recall.
+
+\item It is strongly recommended that lots of thought is placed
+into the settings of the CC timers. Our general recommendation is
+that timers for phones should be set shorter than those for trunks.
+The reason for this is that it makes it less likely for a link in
+the middle of a network to cause CC to fail.
+
+\item CC can potentially be a memory hog if used irresponsibly. The
+following are recommendations to help curb the amount of resources
+required by the CC engine. First, limit the maximum number of
+CC requests in the system using the cc\_max\_requests option in
+ccss.conf. Second, set the cc\_offer\_timer low for your callers. Since
+it is likely that most calls will not result in a CC request, it is
+a good idea to set this value to something low so that information
+for calls does not stick around in memory for long. The final thing
+that can be done is to conditionally set the cc\_agent\_policy to
+"never" using the CALLCOMPLETION dialplan function. By doing this,
+no CC information will be kept around after the call completes.
+
+\item It is possible to request CCNR on answered calls. The reason
+for this is that it is impossible to know whether a call that is
+answered has actually been answered by a person or by something
+such as voicemail or some other IVR.
+
+\item Not all channel drivers have had the ability to set CC config
+parameters in their configuration files added yet. At the time of
+this writing (2009 Oct), only chan\_sip has had this ability added, with
+short-term plans to add this to chan\_dahdi as well. It is
+possible to set CC configuration parameters for other channel types,
+though. For these channel types, the setting of the parameters can
+only be accomplished using the CALLCOMPLETION dialplan function.
+
+\item It is documented in many places that generic agents and monitors
+can only be used for phones. In most cases, however, Asterisk has no
+way of distinguishing between a phone and a trunk itself. The result
+is that Asterisk will happily let you violate the advice given and
+allow you to set up a trunk with a generic monitor or agent. While this
+will not cause anything catastrophic to occur, the behavior will most
+definitely not be what you want.
+
+\item At the time of this writing (2009 Oct), Asterisk is the only
+known SIP stack to write an implementation of
+draft-ietf-bliss-call-completion-04. As a result, it is recommended
+that for your SIP phones, use a generic agent and monitor. For SIP
+trunks, you will only be able to use CC if the other end is
+terminated by another Asterisk server running version 1.8 or later.
+
+\item If the Dial application is called multiple times by a single
+extension, CC will only be offered to the caller for the parties called
+by the first instantiation of Dial.
+
+\item If a phone forwards a call, then CC may only be requested for
+the phone that executed the call forward. CC may not be requested
+for the phone to which the call was forwarded.
+
+\item CC is currently only supported by the Dial application. Queue,
+Followme, and Page do not support CC because it is not particularly
+useful for those applications.
+
+\item Generic CC relies heavily on accurate device state reporting. In
+particular, when using SIP phones it is vital to be sure that device
+state is updated properly when using them. In order to facilitate proper
+device state handling, be sure to set callcounter=yes for all peers and
+to set limitonpeers=yes in the general section of sip.conf
+
+\item When using SIP CC (i.e. native CC over SIP), it is important that
+your minexpiry and maxexpiry values allow for available timers to run
+as little or as long as they are configured. When an Asterisk server
+requests call completion over SIP, it sends a SUBSCRIBE message with
+an Expires header set to the number of seconds that the available
+timer should run. If the Asterisk server that receives this SUBSCRIBE
+has a maxexpiry set lower than what is in the received Expires header,
+then the available timer will only run for maxexpiry seconds.
+
+\item As with all Asterisk components, CC is not perfect. If you should
+find a bug or wish to enhance the feature, please open an issue on
+https://issues.asterisk.org. If writing an enhancement, please be sure
+to include a patch for the enhancement, or else the issue will be
+closed.
+
+\end{itemize}
+
+\section{Simple Example of generic call completion}
+
+The following is an incredibly bare-bones example sip.conf
+and dialplan to show basic usage of generic call completion.
+It is likely that if you have a more complex setup, you will
+need to make use of items like the CALLCOMPLETION dialplan
+function or the CC\_INTERFACES channel variable.
+
+First, let's establish a very simple sip.conf to use for this
+
+\begin{verbatim}
+[Mark]
+context=phone_calls
+cc_agent_policy=generic
+cc_monitor_policy=generic
+;We will accept defaults for the rest of the cc parameters
+;We also are not concerned with other SIP details for this
+;example
+
+[Richard]
+context=phone_calls
+cc_agent_policy=generic
+cc_monitor_policy=generic
+\end{verbatim}
+
+Now, let's write a simple dialplan
+
+\begin{verbatim}
+[phone_calls]
+
+exten => 1000,1,Dial(SIP/Mark,20)
+exten => 1000,n,Hangup
+
+exten => 2000,1,Dial(SIP/Richard,20)
+exten => 2000,n,Hangup
+
+exten => 30,1,CallCompletionRequest
+exten => 30,n,Hangup
+
+exten => 31,1,CallCompletionCancel
+exten => 31,n,Hangup
+\end{verbatim}
+
+\begin{itemize}
+\item Scenario 1:
+Mark picks up his phone and dials Richard by dialing 2000. Richard is
+currently on a call, so Mark hears a busy signal. Mark then hangs up,
+picks up the phone and dials 30 to call the CallCompletionRequest
+application. After some time, Richard finishes his call and hangs up.
+Mark is automatically called back by Asterisk. When Mark picks up his
+phone, Asterisk will dial extension 2000 for him.
+
+\item Scenario 2:
+Richard picks up his phone and dials Mark by dialing 1000. Mark has stepped
+away from his desk, and so he is unable to answer the phone within the
+20 second dial timeout. Richard hangs up, picks the phone back up and then
+dials 30 to request call completion. Mark gets back to his desk and dials
+somebody's number. When Mark finishes the call, Asterisk detects that Mark's
+phone has had some activity and has become available again and rings Richard's
+phone. Once Richard picks up, Asterisk automatically dials exteision 1000 for
+him.
+
+\item Scenario 3:
+Much like scenario 1, Mark calls Richard and Richard is busy. Mark hangs up,
+picks the phone back up and then dials 30 to request call completion. After
+a little while, Mark realizes he doesn't actually need to talk to Richard, so
+he dials 31 to cancel call completion. When Richard becomes free, Mark will
+not automatically be redialed by Asterisk.
+
+\item Scenario 4:
+Richard calls Mark, but Mark is busy. About thirty seconds later, Richard decides
+that he should perhaps request call completion. However, since Richard's phone
+has the default cc\_offer\_timer of 20 seconds, he has run out of time to
+request call completion. He instead must attempt to dial Mark again manually. If
+Mark is still busy, Richard can attempt to request call completion on this second
+call instead.
+
+\item Scenario 5:
+Mark calls Richard, and Richard is busy. Mark requests call completion. Richard
+does not finish his current call for another 2 hours (7200 seconds). Since Mark
+has the default ccbs\_available\_timer of 4800 seconds set, Mark will not be
+automatically recalled by Asterisk when Richard finishes his call.
+
+\item Scenario 6:
+Mark calls Richard, and Richard does not respond within the 20 second dial timeout.
+Mark requests call completion. Richard does not use his phone again for another
+4 hours (144000 seconds). Since Mark has the default ccnr\_available\_timer
+of 7200 seconds set, Mark will not be automatically recalled by Asterisk when
+Richard finishes his call.
+\end{itemize}