You should consider certain product runtime behaviors when
you are writing Session Initiation Protocol (SIP) applications.
Container may accept non-SIP URI schemes
The
SIP container will not reject a message if it doesn't recognize the
scheme in the request Uniform Resource Indicator (URI) because the
container cannot know which URI schemes are supported by the applications.
SIP elements may support a request URI with a scheme other than sip or sips,
for example, the pres: scheme has a particular meaning
for presence servers, but the container does not recognize it. It
is up to the application to determine whether to accept or to reject
a specific scheme. SIP elements may translate non-SIP URIs using any
mechanism available, resulting in SIP URIs, SIPS URIs, or other schemes,
like the tel URI scheme of RFC 2806 [9].
For transitioning users: When a SIP application sends a request
to a SIP URI over Transport Layer Security (TLS) in version 6.1, the
request URI scheme changes from "sip" to "sips." In version 7.0, the
scheme does not change. You can reverse the new behavior in version
7.0 by changing the application code. With a "sips" URI, the behavior
remains the same after upgrading from version 6.1 to 7.0. See the
information center topic Premigration considerations for more information.
trns
Directing requests
in a multiple-container environment
In a multiple-container
environment (SIP proxy plus SIP containers), when your application
sends a request intended initially to be sent externally but later
received, it should use the host and ports of the front-most load
balancing element (either an IP sprayer for multiple SIP proxies,
or the SIP proxy if only one exists). If the application uses the
host name of a container instead of the front-most element, the request
may be lost in the event of a failure.
For example, an application
sends an INVITE request to itself, but the request must pass through
an external accounting system through a pushed Route header. The application
should set the INVITE request's URI to the host and port of the foremost
element to ensure that failover occurs. The request will be routed
to the accounting system via the pushed Route, and then sent back
to the front load balancing element for processing.
Invoking session listener events
SipSessionListener
and SipApplicationSessionListener events are invoked only if an application
requests the corresponding session object. You do this by using in
your application the method shown in
Methods that invoke session listener events.
Table 1. Methods that invoke session listener
events. This table lists the methods
that invoke session listener events.
Event |
Method |
SipSessionListener |
getSession() |
SipApplicationSessionListener |
getApplicationSession() |
Session activation and passivation
During normal
operation, this product never migrates a session from one server to
another. Session migration occurs only as a result of a server failure.
Therefore the SipSessionActivationListener method's passivation callback
is never invoked. However, the activation callback is invoked when
a failure forces session failover to a different server.
External resources
If a SIP application performs
intensive I/O or accesses an external database, it may be blocked
for several milliseconds. If possible, use asynchronous APIs for these
resources . Under stress, a blocked SIP application may trigger a
Request Timeout or re-transmission.
SIP application attributes
Avoid hanging large
objects or BLOBs as SIP Session attributes (via SIPSession.setAttribute
API). This may damage the overall performance when combined with high
availability (HA). The same recommendation applies for SIPApplicationSession.setAttribute.
In most cases, the large object can be replaced by several simple
or composed strings.