• United States

Take a SIP, but don’t bury H.323

Jun 07, 20043 mins

During my 20 years in IT, I have seen the premature death notices of many technologies that are still very much alive. I have read the obituaries of SNA, token ring, ISDN and COBOL. Analysts have invited me to witness the demise of the mainframe, LAN bridging and IPv4. And yet, each one is still in use today. Some have taken on new identities (LAN bridging is now called Layer 2 switching), while others still are functioning in their original wrappers more than 10 years after the proclamation of their death.

So it was with amusement that I recently heard some analysts proclaim the death of H.323. Both beloved and hated, H.323 is the epitome of the two-faced devil for VoIP. On the one hand, H.323 has been the enabler of almost all IP-based telephony implementations in use. On the other hand, H.323 has been the biggest barrier to fully deploying true multi-vendor IP telephony.

H.323’s failing is partly because of one of its original benefits – its flexibility. Vendors could manipulate the H.323 specification so much that it was possible to have an “H.323-compliant” device that could not interoperate with another vendor’s “H.323-compliant” device. H.323 also suffered from performance issues. Vendors could develop their own proprietary algorithms that provided better voice quality and less latency.

To overcome these issues, the IETF developed Session Initiation Protocol (SIP). SIP provides the same signaling functionality as H.323 and can be used for session initiation, call setup, call tear-down, call routing and error handling.

Because H.323 originally was designed for multimedia applications, it is complex and requires multiple messages to establish a communications session. SIP was developed primarily for VoIP and requires less overhead. SIP can set up a call with a minimum of three messages, thus increasing performance over H.323.

Being a text-based protocol, SIP implementations are easier to develop and debug. Implementing multi-vendor interoperability is proving to be much easier with SIP. Unlike H.323, SIP-compliant devices communicate with one another.

SIP is not perfect. Because it is User Datagram Protocol-based and uses explicit static addressing, it cannot traverse network address translation gateways unless Simple Traversal of UDP through NAT, NAT probes or other SIP-aware mechanisms are implemented. Despite this flaw, SIP is a major advance over H.323 and probably will become the protocol of choice for VoIP call signaling. However, to call H.323 dead is premature.

Last July, Gartner predicted that by the end of 2006, 50% of IP enterprise telephony systems will support SIP. That means that 50% of the implementations will use non-SIP options such as H.323. Most companies will want to maximize their investment, so these non-SIP systems likely will remain in place for at least five years. This means H.323 will be alive and kicking until at least 2011.

H.323 is as “dead” as SNA, ISDN or COBOL. While it may no longer be part of the overall VoIP strategic direction, it will continue to be used well into the next decade.