Some general questions about H.323 and SIP

[Stefan]

Hi erverybody!

My name is Stefan, I'm from Germany and at the moment I'm writing my diploma thesis about VoIP over WLAN / WiFi. While studying some Websites and Books I found out that most people think SIP is best fitting for VoIP. Finally I found this Website and it made me happy to see some arguments form an other point of view because most of these websites don't explain why SIP is much more better than H.323. I jut wanted to ask some questions and tell you my point of view. Feel free to comment, critcise or correct my theories, questions and statements.

1. Nowadays most users (at home or with mobile phones) are using SIP for Voice. H.323 is mostly used in corporative networks and for video conferencing. I've read that SIP is growing in these areas too. But in the HD-video area vendors like Cisco only using H.323. Is this all true or are there any differences what you believe or know?

2. SIP nowadays is the most used VoIP protocol because it has the ability to add services and extensions easily . It's easier for programmers to learn it and create new applications. If you just need a voice capable ip-phone a vendor has less programming efford because they only need to add the voice extensions. H.323 has to be compatible to every other H.323 application. I read this somewhere in the internet and there were no references so I'm not sure if this really is true.

3. Nowadays SIP and H.323 have a lot of things in common and most SIP/H.323 can do H.323/SIP can do also. So what are the features of SIP and H.323 that stand out the most ? Which advantages and disadvantages are relevant regarding todays usage?

4. The future is NGN but SIP and H.323 are not fitting perfect into this. So H.325 or an alternative protocol is going to take over the market some day. How long will this take? Maybe 10 15 years or will there a hype about this? What do you think?

5.If I want to implement a VoIP structure for a company with 100 VoIP users. The users will have IP-Phones and WiFi-IP-Phones and maybe some will also have Softphones. The users will call each other in that company and also people located anywhere in the world (whatever they are using VoIP or ISDN or something else). What benefits would H.323 deliver which SIP can't. I think maybe the interoperability with the PSTN would be better with H.323. If the Users want to do video calls this would also be an advantage of H.323. What else could turn the balance to H.323? Maybe there would be more different Vendors/Products for SIP, so maybe the devices would be cheaper (because they haven't got the features of H.323 devices)?

Thank you for any response. Even it's only for one Question/Statement. I'm new at this topic so I just want to get in contact with developers and users who have several years of knowledge and practise in VoIP/SIP/H.323 and want to share their experience.

Thanks a lot!

Stefan

[paulej]

My name is Stefan, I'm from Germany and at the moment I'm writing my diploma thesis about VoIP over WLAN / WiFi. While studying some Websites and Books I found out that most people think SIP is best fitting for VoIP. Finally I found this Website and it made me happy to see some arguments form an other point of view because most of these websites don't explain why SIP is much more better than H.323. I jut wanted to ask some questions and tell you my point of view. Feel free to comment, critcise or correct my theories, questions and statements.

Yeah, most will just argue that SIP is better because they were told so, but have no actual hands-on experience. Truth is, the work required to implement either SIP or H.323 is about the same. Back in 2000, Jeff Pulver called H.323 a dinosaur, yet both H.323 and SIP are about the same age. About the same time, a group of companies (including Nortel) argued that SIP should be adopted by 3GPP, which was fine, but they argued against H.323 using arguments that were not entirely founded. They made the most fundamental mistake of taking the most complex call flow and comparing it against the simplest SIP flow. Now look at the typical IMS flow from 3GPP-- it's complex.

Ten years later, we still don't have very good interoperability between SIP systems. There are about as many flavors of SIP as there are implementations.

There has been a strange "marketing engine" behind SIP forever. There were claims that SIP is so much simpler, it will do all kinds of things that were never possible before, etc. What can it do in practice? It has proven to be nothing more than a replacement for the PSTN with functionality that's virtually the same. It's good that we moved from PSTN to IP, but what new capabilities did end users get? Not much. There is video, but we could have done that with H.323 even better.

1. Nowadays most users (at home or with mobile phones) are using SIP for Voice. H.323 is mostly used in corporative networks and for video conferencing. I've read that SIP is growing in these areas too. But in the HD-video area vendors like Cisco only using H.323. Is this all true or are there any differences what you believe or know?

Cisco uses both protocols. Cisco's Telepresence systems use SIP with "special sauce". The behavior was documented and published for the benefit of other vendors, but the non-standard behavior was necessary for reasons, not the least of which is that SIP does not define how to do video. Sure, one can open a video stream, but there's much more to videoconferencing than just sending a video stream.

2. SIP nowadays is the most used VoIP protocol because it has the ability to add services and extensions easily . It's easier for programmers to learn it and create new applications. If you just need a voice capable ip-phone a vendor has less programming efford because they only need to add the voice extensions. H.323 has to be compatible to every other H.323 application. I read this somewhere in the internet and there were no references so I'm not sure if this really is true.

I do not know what the percentage is, exactly. I do know that Skype owns 14% of the market.

H.323 definitely dominates the videoconferencing market and has a healthy share of the international long distance market.

SIP is widely used for residential VoIP services and it's gaining momentum in enterprise VoIP as a replacement for the legacy PSTN. It's also gaining share in the mobile market (3GPP/IMS) where it was slated for use 10 years ago. However, end users do not really see that. This is really very much "behind the scenes". End users just get voice on their mobile devices like before. Perhaps it's not actually deployed all the way to the handset, since bandwidth issues have prevented that from being very usable in the past. Since I'm not actively working in this area, I've not followed-up to see. But, I would bet it is still an issue.

3. Nowadays SIP and H.323 have a lot of things in common and most SIP/H.323 can do H.323/SIP can do also. So what are the features of SIP and H.323 that stand out the most ? Which advantages and disadvantages are relevant regarding todays usage?

We've tried to capture those in the H.323 vs SIP page. At the end of the day, both protocols do pretty much the same thing. H.323 still has a strength in videoconferencing, whereas SIP does have an advantage in being lighter-weight in simple terminal devices, in theory. (I say that, because in order to build a terminal that does all of the proper signaling to be a robust system, you need to implement a LOT of RFCS and the complexity goes up as a result.)

4. The future is NGN but SIP and H.323 are not fitting perfect into this. So H.325 or an alternative protocol is going to take over the market some day. How long will this take? Maybe 10 15 years or will there a hype about this? What do you think?

I question whether the future is truly NGN. NGN started out with some good ideas, but what I've seen it turn into is an effort by some to more-or-less re-invent the Internet. Why do that? Do users want that? I think NGN introduces some good ideas that might be borrowed, but I'm skeptical that it will replace the Internet.

We are defining H.325 to work with the NGN, but we definitely want it to work on the Internet, too. The concept behind H.325 is to create a very different kind of multimedia system that addresses the shortcomings of both H.323 and SIP. Of particular interest to me, I want a system that enables application developers to build all kinds of multimedia applications that I can use seamlessly in my communication with another person.

5.If I want to implement a VoIP structure for a company with 100 VoIP users. The users will have IP-Phones and WiFi-IP-Phones and maybe some will also have Softphones. The users will call each other in that company and also people located anywhere in the world (whatever they are using VoIP or ISDN or something else). What benefits would H.323 deliver which SIP can't. I think maybe the interoperability with the PSTN would be better with H.323. If the Users want to do video calls this would also be an advantage of H.323. What else could turn the balance to H.323? Maybe there would be more different Vendors/Products for SIP, so maybe the devices would be cheaper (because they haven't got the features of H.323 devices)?

This is the big question. Most vendors major vendors tend to support both, just "to be safe". One issue with the enterprise is that both H.323 and SIP fall a bit short on features needed for enterprise users. As such, a proprietary line-side protocol is probably better where features like park/pick-up, "boss/secretary", and other features are needed.

For deployments where phones are fairly autonomous (i.e., no central "iPBX" controlling their every function), then both SIP and H.323 will work. For voice, SIP really does have the "mind share" and so that might make the most sense. It's not really a technical decision, but "following the crowd charging over the cliff" kind of decision

If video is the primary concern, then H.323 might be the better choice. Another advantage of H.323 now that we have NAT/FW specs approved is that H.323 can more easily get through NAT/FW devices without pushing media into a media relay device (see H.460.23/H.460.24). Those do require well-behaved NAT devices and they do not work with symmetric NAT, but it's still possible to build large-scale H.323 systems. Check out h323.net: that is operational and people can use it today to use H.323 devices that implement those standards. Presently, only http://www.pacphone.com supports the new specs, since they're so new.

Paul

[Stefan]

Hey Paul.

Thanks a lot for your detailed answer. Some Points made me think a lot

Yeah, most will just argue that SIP is better because they were told so, but have no actual hands-on experience. Truth is, the work required to implement either SIP or H.323 is about the same. Back in 2000, Jeff Pulver called H.323 a dinosaur, yet both H.323 and SIP are about the same age. About the same time, a group of companies (including Nortel) argued that SIP should be adopted by 3GPP, which was fine, but they argued against H.323 using arguments that were not entirely founded. They made the most fundamental mistake of taking the most complex call flow and comparing it against the simplest SIP flow. Now look at the typical IMS flow from 3GPP-- it's complex.

I was searching for an mobile phone which has some VoIP functionality some months ago. In every case I found devices with SIP implemented. This probably regards to the 3GPP. I also read a white paper from nokia where SIP is regarded as smaller and more efficient than H.323 and using other Protocols like MGCP. It's not really comparing SIP and H.323 but talking about how much better SIP is. Maybe you want to have a look at it but its from the year 2003 ( http://www.nokia.com/NOKIA_COM_1/About_ ... r_voip.pdf )

I do not know what the percentage is, exactly. I do know that Skype owns 14% of the market.

H.323 definitely dominates the videoconferencing market and has a healthy share of the international long distance market.

So have you any link where it's stated that Skype has 14% ownage of the market?
I just found out about H.323: "Today, most of the world's VoIP traffic is carried over H.323 networks, with billions of minutes of traffic being carried every month. " But with no proof or references. The site was updated in 2009 the last time maybe it's true maybe not. ( http://www.voip-info.org/wiki/view/H.323 )
I've found out that in Germany only 9% are VoIP calls but they are increasing in the last few years. It's stated by the Bundesnetzagentur (it's the agency of german government to regulate and supervise the telephone internet and mobile market.) The report is written in german so it's probably useless for you. ;(

SIP is widely used for residential VoIP services and it's gaining momentum in enterprise VoIP as a replacement for the legacy PSTN. It's also gaining share in the mobile market (3GPP/IMS) where it was slated for use 10 years ago. However, end users do not really see that. This is really very much "behind the scenes". End users just get voice on their mobile devices like before. Perhaps it's not actually deployed all the way to the handset, since bandwidth issues have prevented that from being very usable in the past. Since I'm not actively working in this area, I've not followed-up to see. But, I would bet it is still an issue.

Ten years ago, I believe the transmission rate wasn't that good as it's nowadays. But 10 years ago, applications don't need as much transmission rate either, I think

First I thougt about that, too, in VoIP general. Because SIP messages take more time travelling around a network because of their readable code and H.323 not because of it's binary code. But nowadays in corporate networks there is gigabit ethernet and it's not that critical I think? Maybe in the mobile market it's a different thing.

This is the big question. Most vendors major vendors tend to support both, just "to be safe". One issue with the enterprise is that both H.323 and SIP fall a bit short on features needed for enterprise users. As such, a proprietary line-side protocol is probably better where features like park/pick-up, "boss/secretary", and other features are needed.

For deployments where phones are fairly autonomous (i.e., no central "iPBX" controlling their every function), then both SIP and H.323 will work. For voice, SIP really does have the "mind share" and so that might make the most sense. It's not really a technical decision, but "following the crowd charging over the cliff" kind of decision

If video is the primary concern, then H.323 might be the better choice. Another advantage of H.323 now that we have NAT/FW specs approved is that H.323 can more easily get through NAT/FW devices without pushing media into a media relay device (see H.460.23/H.460.24). Those do require well-behaved NAT devices and they do not work with symmetric NAT, but it's still possible to build large-scale H.323 systems. Check out h323.net: that is operational and people can use it today to use H.323 devices that implement those standards. Presently, only http://www.pacphone.com supports the new specs, since they're so new.

I'm writing my diploma thesis about some installation in an university of applied sciences. They use H.323 and VoIP and VoIP over WLAN. The VoIP over WLAn isn't really working. Often there are deregistrations and abortions while people are talking with their handhelds. The WLAN is an Meru Networks installation and the VoIP devices are from Innovaphone. The Wireless devices are originally from Ascom but Innovaphone put their Software on them.
It seems that there are too much Access Points for 802.11 b/g but for 802.11 n coverage they are probably needed. We tried to reduce the power they send out their b/g signal but it seems there are more problems than just this one. There is 802.11 traffic in almost every channel and we can't locate the sources. Also there are two sorts of AP's installed and Meru thinks their AP's can't work together correctly. So they think we have to deinstall the cheap ones and install the expensive ones


So I've heard there is an business free day on monday. So have a nice 4 day weekend!
Thanks for your reply.

Stefan

[paulej]

I was searching for an mobile phone which has some VoIP functionality some months ago. In every case I found devices with SIP implemented. This probably regards to the 3GPP. I also read a white paper from nokia where SIP is regarded as smaller and more efficient than H.323 and using other Protocols like MGCP. It's not really comparing SIP and H.323 but talking about how much better SIP is. Maybe you want to have a look at it but its from the year 2003 ( http://www.nokia.com/NOKIA_COM_1/About_ ... r_voip.pdf )

I guy I work with at Nokia also said, "SIP sucks. It really sucks. But, it's what we have."

I think most who work with SIP for any length of time will tell you it has problems. But, for the mobile carrier market, SIP is it, because 3GPP adopted it and that's what the operators are going with. Having said that, I've been told about several RFPs for an H.323-based mobile phone application from carriers and/or companies that want to provide "over the top" mobile video calling services. I think in all cases, people are looking to H.323 for video.

So have you any link where it's stated that Skype has 14% ownage of the market?
I just found out about H.323: "Today, most of the world's VoIP traffic is carried over H.323 networks, with billions of minutes of traffic being carried every month. " But with no proof or references. The site was updated in 2009 the last time maybe it's true maybe not. ( http://www.voip-info.org/wiki/view/H.323 )
I've found out that in Germany only 9% are VoIP calls but they are increasing in the last few years. It's stated by the Bundesnetzagentur (it's the agency of german government to regulate and supervise the telephone internet and mobile market.) The report is written in german so it's probably useless for you. ;(

See the chart on the front of http://www.h323forum.org. That graph was correct as of 2006. SIP had very little traction at that time and, where it did, interconnection was always through the PSTN.

Since then, more emphasis has been put on SIP (people going along with the hype). Skype has continued to gain momentum. I mis-spoke about the 14%: it's actually 12%. Here's a reference: http://www.channelweb.co.uk/crn/comment ... ass-market.

Ten years ago, I believe the transmission rate wasn't that good as it's nowadays. But 10 years ago, applications don't need as much transmission rate either, I think

Bandwidth is really still a problem. Here in the US, you're lucky to get good 3G coverage and, when you do have it, the jitter and delay are horrible. I ran Skype over 3G once... it wasn't worth it. I have never had such a horrible phone call before. Perhaps the carriers can do better by routing traffic through SBCs under their control, but I don't think they can provide the level of quality people expect, yet.

First I thougt about that, too, in VoIP general. Because SIP messages take more time travelling around a network because of their readable code and H.323 not because of it's binary code. But nowadays in corporate networks there is gigabit ethernet and it's not that critical I think? Maybe in the mobile market it's a different thing.

The size of the messages are not really an issue. People want to send Mbps of video through, so a few KB of signaling is, in my opinion, not an issue. In fact, we plan to use XML in H.325. XML would bring a level of consistency to all signaling and application interfaces. Had XML been around in 1995 when work on SIP got started, perhaps it might have been chosen. Now we have to deal with all kinds of syntax headaches with SIP. Every header line has a different syntax from the next. Then SDP is a whole different syntax, with each of its lines having a unique syntax. It really makes SIP a lot more complex to properly process than some folks think... until they try to do it. (I can build a half-assed implementation pretty quickly, but a robust implementation requires a lot of work.)

I'm writing my diploma thesis about some installation in an university of applied sciences. They use H.323 and VoIP and VoIP over WLAN. The VoIP over WLAn isn't really working. Often there are deregistrations and abortions while people are talking with their handhelds. The WLAN is an Meru Networks installation and the VoIP devices are from Innovaphone. The Wireless devices are originally from Ascom but Innovaphone put their Software on them.

Do you have any QoS features enabled on the WLAN? Without that, it seems you would likely have problems when more users are added. Further, you need really good coverage. Even in my own house, the bit-rate drops off significantly from one side of the house to the other due to a drop in signal strength. That said, my wife does voice/video over the wireless network daily in excess of 256Kbps and it works perfectly.

It seems that there are too much Access Points for 802.11 b/g but for 802.11 n coverage they are probably needed. We tried to reduce the power they send out their b/g signal but it seems there are more problems than just this one. There is 802.11 traffic in almost every channel and we can't locate the sources. Also there are two sorts of AP's installed and Meru thinks their AP's can't work together correctly. So they think we have to deinstall the cheap ones and install the expensive ones

I'm really not an expert in wirless LAN area, so I can't provide good guidance. But, I can say this definitely has focus. There are folks I work with who are focused on this problem. Whether it's a new standard for QoS, deployment guidelines, or something else, I'm really not sure. I can only speculate, myself.

So I've heard there is an business free day on monday. So have a nice 4 day weekend!

Yes, it is. And I might even get away from the computer for a change

Thanks!
Paul

[d.kochmashev]

One issue with the enterprise is that both H.323 and SIP fall a bit short on features needed for enterprise users. As such, a proprietary line-side protocol is probably better where features like park/pick-up, "boss/secretary", and other features are needed.

Paul, do you think that H.450 is not suitable for that? Can you share your point of view with us?

[paulej]

One issue with the enterprise is that both H.323 and SIP fall a bit short on features needed for enterprise users. As such, a proprietary line-side protocol is probably better where features like park/pick-up, "boss/secretary", and other features are needed.

Paul, do you think that H.450 is not suitable for that? Can you share your point of view with us?

The problem is that each individual H.450 feature requires supporting software. The more features you add to the endpoint, the more complex the endpoint becomes.

For that reason, where features like this are needed and often used, it's simpler to have a device control protocol in use (e.g., H.248 or MGCP).

Paul

[Stefan]

Hi everybody.

Bandwidth is really still a problem. Here in the US, you're lucky to get good 3G coverage and, when you do have it, the jitter and delay are horrible. I ran Skype over 3G once... it wasn't worth it. I have never had such a horrible phone call before. Perhaps the carriers can do better by routing traffic through SBCs under their control, but I don't think they can provide the level of quality people expect, yet.

Here in Germany most regions are covered good. Only a few countrysides the coverage isn't very good. Maybe LTE will change this. (Just two or three weeks ago there was a auction for frequencies for LTE).
But I don't know anything about delay or jitter.

Do you have any QoS features enabled on the WLAN? Without that, it seems you would likely have problems when more users are added. Further, you need really good coverage. Even in my own house, the bit-rate drops off significantly from one side of the house to the other due to a drop in signal strength. That said, my wife does voice/video over the wireless network daily in excess of 256Kbps and it works perfectly.

Yes QoS is enabled and the coverage is very good. Meru thinks its too good for 802.11 b/g. ( The APs are only 5-10 meters away from each other).
The last few days we are testing the whole scenario here at our office. We just use one AP and two mobile devices and the problem is still there. The WLAN coverage iss good (the phones are just 2 or 3 meters away from the AP) and nobody is logged in to the network except for the two phones. It seems there is a problem with the communication between access point and phone. There is a packet loss rate of 20% and the whole amount of failures cause the registration problems.

See the chart on the front of http://www.h323forum.org. That graph was correct as of 2006. SIP had very little traction at that time and, where it did, interconnection was always through the PSTN.

Since then, more emphasis has been put on SIP (people going along with the hype). Skype has continued to gain momentum. I mis-spoke about the 14%: it's actually 12%. Here's a reference: http://www.channelweb.co.uk/crn/comment ... ass-market.

As written in my Email some hours ago you have to pay for the new charts so I have to take the old ones

I've found a website with a lot of carriers listed. There is also written which protocols they use. Maybe you are interested in that.

http://www.voip-catalog.com/voip_protocols.html

It seems SIP and H.323 approximately have the same usage by carriers. You can see that there's a little more usage of SIP.

If video is the primary concern, then H.323 might be the better choice. Another advantage of H.323 now that we have NAT/FW specs approved is that H.323 can more easily get through NAT/FW devices without pushing media into a media relay device (see H.460.23/H.460.24). Those do require well-behaved NAT devices and they do not work with symmetric NAT, but it's still possible to build large-scale H.323 systems. Check out h323.net: that is operational and people can use it today to use H.323 devices that implement those standards. Presently, only http://www.pacphone.com supports the new specs, since they're so new.

I've read some things about NAT and Firewalls and VoIP. Some say with H.323 its more easier others say its easier with SIP becuase of the SIP Proxy. I don't know much about NAT and Firewalls. The only thing I know is that H.323 is working at the university we have installed it and they use NAT and firewalls. But I think it would run there with SIP because we tested a SIP Softphone too and it worked without problems.

Have a nice day

Stefan

[paulej]

I've read some things about NAT and Firewalls and VoIP. Some say with H.323 its more easier others say its easier with SIP becuase of the SIP Proxy. I don't know much about NAT and Firewalls. The only thing I know is that H.323 is working at the university we have installed it and they use NAT and firewalls. But I think it would run there with SIP because we tested a SIP Softphone too and it worked without problems.

Both SIP and H.323 present NAT/FW problems. Some have argued that SIP is easier for a few reasons:
1) UDP is often used for signaling, so devices can send a REGISTER message over that port to keep the port open on the NAT/FW device to enable incoming calls as a bi-product
2) Historically (and even today), the only media flow usually established is a single voice flow
2a) The UDP port advertised in the SDP is often ignored entirely
2b) The UDP port for voice is determined when the endpoint starts sending media to the service provider's allocated voice port
3) Symmetric ports are used for transmitting and receiving audio, so there's nothing special required to get media through the NAT/FW

Some of these properties are also true of H.323, but not all. H.323 devices use both UDP and TCP for signaling (RAS and "call signaling"). This makes it more challenging to get calls through the NAT/FW. This could be addressed using something like H.460.17 or perhaps H.323 Annex E. Another issue for H.323 is that voice is often not the only flows. It is not unusual to have voice, video, far-end camera control (FECC), etc.

We're now starting to see more complexity introduced on the SIP side. Things like MSRP, BFCP, etc. introduce more NAT/FW issues, so what used to work for only voice will not work anymore. This has been a motivator for introducing complex procedures like ICE.

Paul

[Stefan]

Hi Paul

if found this statement in an other forum

SIP itself isn't really workable behind NAT routers, and that has led to some kludges that have helped it along. Not easily, and not as well as if it had been designed to handle being handed out a private IP in the first place, but so far the kludges hold up.

would you agree with that? Or do you think this is not true?

Stefan

[paulej]

Hi Paul

if found this statement in an other forum

SIP itself isn't really workable behind NAT routers, and that has led to some kludges that have helped it along. Not easily, and not as well as if it had been designed to handle being handed out a private IP in the first place, but so far the kludges hold up.

would you agree with that? Or do you think this is not true?

Stefan

Yeah, that's true. There's nothing in SIP itself that facilitates being behind a NAT device. But, using "kludges" (which I suspect are some of the things I mentioned above), it's workable. This does assume that the SBC-type device in the public network and the endpoints behind the NAT make the same kinds of assumptions (e.g., symmetric port usage).

Paul