Response to "Two VoIP protocols compared"

From: []On Behalf Of Paul E. Jones
Sent: Tuesday, January 01, 2002 10:19 AM
Subject: [h323forum] Extremely misinformed report


I was referred to this report written by Phil Hochmuth:

This report makes statements that are entire false. The reported obviously did not contact anybody from the H.323 community in order to draft the report. If he had, he would have heard quite a different story.

Let me comment on a few statements:

``The origins of H.323 and Session Initiation Protocol are about as disparate as the histories of the World Wide Web and Ma Bell's century-old network.''

This is only half true. H.323 was drafted to be a voice, video, data conferencing protocol for LANs, whereas SIP was originally intended to be a "point to point" version of the Session Announcement Protocol (SAP). Both of these protocols have changed since their original inception.

`T`he H.323 protocol came on the scene in the mid-'90s as a transmission and session setup protocol for videoconferencing over ISDN networks.''

This is entirely false. H.323 was developed with packet-based videoconferencing in mind. The only thing borrowed from the ISDN world were some of the basic messages like "Setup", "Connect", etc. Those parallel the same messages as in SIP, "INVITE" and "200". Additionally, the bulk of H.323 is actually new and created specifically for the application of packet-based multimedia conferencing. This protocol cannot be used in its original or current form for ISDN networks, so the authors statement is bogus.

``SIP, on the other hand, was borne out of the Internet Engineering Task Force (IETF), the governing body that oversees development and standardization of all Internet technologies.''

The IETF is only one organization that standardizes protocol for the Internet and it is certainly not the "governing body". In fact, the W3C, ITU, and ETSI (among others) all have active groups developing protocols for use in the IP networks. Most everybody recognizes the IETF as the right place for transport issues, but everybody has an opinion about where everything else should be done.

``H.323 networks rely on a central server or device for tasks such as call processing and applications, with "dumb devices" (phones) typically deployed as endpoints in the network. SIP operates under a more decentralized network scheme, with some intelligence pushed out to "smart" endpoints, such as SIP phones, PCs and wireless devices that have the intelligence to run applications.''

This is also untrue. Both H.323 and SIP can operate with or without centralized devices and centralized intelligence. H.323 and SIP are both considered "intelligent" endpoints and neither are "dumb" phone protocols-- perhaps the author was thinking of MGCP or H.248, which are device control protocols.

``H.323 and SIP are also structured differently. As a text-based protocol, SIP runs on top of IP. With H.323, voice data is encapsulated into IP packets. It is more complex than SIP, making it harder for network managers to troubleshoot and more complex for application developers to work with, experts say.''

The second and third sentences here make no sense. The encoding of messages has nothing to do with "running on top of IP" or being "encapsulated into IP packets". Both H.323 and SIP are considered "application protocols". Indeed, SIP uses a textual encoding for its messages, whereas H.323 encodes things in binary. However, it should also be pointed out that many, many protocols used within the Internet are defined as binary protocols and are done so for both efficiency in processing time and bandwidth utilization on the wire. As for management, it is true that H.323 requires decoding before it can be viewed by the "administrator", but tools are abundant and most products come with diagnostic capabilities. In fact, the professional tool used to snoop on the network for SIP messages will probably also support decoding of H.323 messages. One thing is certain, though: reading network traces is not the best way to manage a communication system. H.323 and SIP equipment should both provide the administrator with the ability to see information necessary to diagnose problems. Also, the claim that the binary encoding is more complex for application developers is a comment only stated by SIP developers and the "pro-text" group-- there is also a large "pro-binary" group. H.323 developers are quite comfortable with the binary encoding. In fact, very little time is spent dealing with the "binary" encoding in H.323. In every H.323 application, there is a layer that will decode the message and from that point forward, the application only deals with data structures. This makes it no different than any other type of application. What's more, the same is true for SIP. There is a layer that will decode the text messages and the application will only deal with data structures. In both H.323 and SIP applications, the bulk of the work is in higher-layer processing and the actual format of the messages is an extremely small issue.

``H.323 was designed for ISDN videoconferencing, not voice over IP, says Tom Valovic, an analyst with IDC. The IETF engineers who crafted SIP always had IP voice in mind.''

Tom is ill-informed. H.323 was not designed for ISDN. Perhaps he is thinking of H.320, which is an entirely different protocol which was designed for video conferencing over ISDN. H.323 was drafted with IP in mind. In fact, H.323 was the first standards-based protocol to adopt the use of RTP for media transmission: the same standard used for transporting media in SIP.

"SIP is more in the next-generation space. H.323 is more of a bridge between older and newer approaches," he says.

I see both of these protocols taking us into the next generation. There are strengths and weaknesses with each protocol, but I don't see H.323 as only serving as a bridge. The fact is, SIP still falls short in many areas. At some point in the future, SIP may catch up to H.323 in terms of capability and completeness. However, SIP does have advantages in areas outside of telephony, such as Instant Messaging. H.323 actually supports text messaging (done to support the hearing impaired). However, SIP is more appropriate for the Instant Messaging application market. Telephony, on the other hand, is a whole other story: H.323 is the clear leader and SIP has some catching up to do in terms of protocol maturity and deployment.

Best Regards,
Paul E. Jones
Editor, Recommendation H.323v4
Rapporteur, ITU-T Question 2 / Study Group 16