- This topic has 21 replies, 1 voice, and was last updated 18 years, 1 month ago by Tran.
31st January 2002 at 17:31 #21170Tom HowardGuest
I know it’s an old question, but I’m interested in people’s views on whether SIP or H.323 will emerge as the dominant signalling protocol for multimedia.
It was written in 2000, and concluded that by 2001, one of the protocols will have asserted its dominance. The paper also concluded that the dominant protocol will be SIP.
That doesn’t seem to have become true. I use Netmeeting reugularly, and it does not support SIP; only H.323.
For what they’re worth, here are my thoughts:
- H.323 is incredibly complicated. Defined using ASN.1 and encoded using PER means it’s bit (or perhaps byte) orientated which makes writing a stack difficult.
- SIP seems to be similar to HTTP. Text based and probably much easier to implement in software (although I haven’t tried it yet).
- My guess is that H.323 appeals to bell-heads because of its similarities with protocols that they are comfortable with (ISUP, SS7, Q.931). (I’m a bell-head myself – it isn’t meant to be an insulting term 😉
- H.323 seems to have gained good market penetration (Netmeeting) and SIP clients seem rare.
- Perhaps it’s too late for SIP. That would be a shame. As a software developer, I prefer the look of SIP.
I’d be very interested in anyone else’s thoughts or corrections.5th February 2002 at 23:01 #21171TaaGuest
You are right that H.323 is complicated.However,if you look at the latest versions,like V3 you will find that it is simplified.However,I bet is that SIP will definately become a big standard…I am not betting on it becoming “THE STANDARD”.
My thoughts…18th February 2002 at 15:32 #21172Peter ThomsenGuest
I would put my money on SIP. It has been named the standard for UMTS or 3G mobile. The 2 big names in VoIP, Cisco and Intel are both supporting SIP (or a subset) today. In 3G SIP is in, H323 is out.
Peter.21st February 2002 at 01:31 #21173RGuest
Tom!!! I still owe you that book. 🙂 Very interesting thread you started.
I am pretty protocol agnostic and I believe all will have their place. I believe that MGCP or megaco or some variant thereof will prevail in the carrier community.
SIP will have its place and soon…windows messenger in XP is SIP based which makes it immediately pervasive. SIP definitely lends itself to mobility and anything that relies on intelligent endpoints.
H.323 V3 IMHO did not bring about enough changes to “simplify” the protocol…it still uses ASN.1 notation, but it did bring about annex E and annex G. IMO v4 will have more impact than v3.
At the end of the day, the battle between SIP and H.323 has become more marketing hype and political, but at the end of the day I think SIP will become prevalent. It will still take another 2 years or so to get more traction. Lets keep in mind that a vast majority of VoIP networks out there today rely on H.323. That being said, everyone seems to be relying on these protocol translators to help the adoption of various protocols. There are a ton of issues in this space that I dont even want to touch right now.
ITU standards are just too slow to bring about change. IETF is just plain faster. SIP is already on bis06.
R B21st February 2002 at 09:33 #21174Tom HowardGuest
Long time – no hear!
As a software developer, I certainly prefer SIP. I am writing an H.225 protocol stack at the moment. It’s going very well. But, before I could start, I needed to understand ASN.1 and PER X.691 [why PER, just to save a few bits here and there 🙁 ]. It has taken weeks to get my head round them. Although they are impecible in their logic, I don’t believe they were written by people who will actually use and implement them. IETF recommendations are far easier to implement.
p.s. Looking forward to seeing the book. Is it out yet?21st February 2002 at 22:16 #21175TaaGuest
Cisco is not actually using SIP.They have their own protocol call “SKINNY”.Although this is a SIP based protocol,but it has it’s own flavors.22nd February 2002 at 00:21 #21176RommelGuest
Taa actually cisco is…
Skinny isnt even close to SIP…its more like MGCP. SIP is in most of the IOS gateways. Skinny is only done on AVVID.18th April 2002 at 07:57 #21177AlexanderGuest
Tom, how many time you think is needed to write H.323 protocol stack for basic VoIP (including RTP, H.225, H.245)?
Cisco IP-Phones 7940, 7960 supports Skinny (native protocol) and there are SIP and (maybe not yet publicly availbale) MGCP downloadable s/w versions. Cisco CallManager supports H.323, Skinny, MGCP (no SIP).
PS. I enjoyed “bell-head” term 🙂18th April 2002 at 10:04 #21178Tom HowardGuest
We wrote as much as we needed, which was a complete H.225.0 stack (including Q.931), about 30% of H.245 and a little T.120. Including research (boy, was there a lot of research!) it took 400 man hours. We didn’t need to implement RTP/RTCP for our application.
The work is now completed, and I have to say that my opinion of H.323 has changed a little since I started this thread.
Yes, H.323 is very complicated. You need to understand ASN.1 and PER encoding (X.691). But, once you have built support for PER, everything else follows on quite nicely. We need to design some protocols for application control, and most likely, we will define them using ASN.1 and encode them using PER, because the framework is now built.
The next job is to research and implement SIP. Well, now I’m not looking forward to that. We need to build something that parses the text messages that SIP generates and will not be able to take advantage of the extensibility and class structures offered (or which can be derived from) H.323’s definition and line coding scheme.
I still think H.323 is too complicated though 🙂19th April 2002 at 12:55 #21179Tom HowardGuest
I found well written comparison of H.323 and SIP at http://www.packetizer.com/iptel/h323_vs_sip/.16th May 2002 at 13:38 #21180AlexanderGuest
Thank you very much for your response. In our development we started with “openh323” stack, but performance issues force us to search for another solution. I think you know, there are open source implementation for SIP and MGCP, but I afraid we’ll start to write our own implementation of H.323 (oh, God!), so your answer is very usefull for me.
Best regards.24th May 2002 at 09:32 #21181Marc HAMMOUCHEGuest
I agree with all, one of the problems between H323 and SIP (or SIP-T)is the complexity of H323.
But I think that the main problem is the interconnection between a VoIP network and a TDM network.
In this case you have two ways :
1- interconnection under ISDN protocol (ITU Q931)
2- interconnection under SS7 protocol (ITU Q7xx)
For case 2 the protocol used as ISUP V3 is really complex with a large number of information elements in messages. The difficulty is to translate these information elements between a ISUP message and a H323 or SIP message for ensure a correct treatment of the call.
It’s necessary to have a complete mapping between the protocol messages, and nobody knows why but H323 is not fully compliant with ISUP (same standard organism – ITU – but no concertation between the teams). At contrary SIP is FULLY compliant with ISUP V3 and that’s the reason it’s a better choice. Think that when you use H323 you add a large delay for the treatment and the translation of the information elements of each message receive from TDM network. With SIP you don’t have this problem.25th May 2002 at 21:27 #21182RommelGuest
Hmm…the H.225 protocol was built with Q931 in mind. I know there is definite guidlines for SS7/C7 interworking with ISDN. Please see Q.699.
Of course SIP-T is fully compliant with ISUP, you encapsulate the whole ISUP message, what is the big deal with that. H.323 was built with TDM in mind, and was built as a way to replicate some of the tdm infrastructure using packet based mechanisms. SIP is an entirely different approach, but the fact that you could have a MIME encoded body in the message allowed the ISUP transport.
BTW…what is ISUP v3? never heard of it…is it one of the red, white or blue specs?
Another note ISUP has parameters, ISDN has information elements. it also sounds like you are a proponent of ISUP transparency. I would like to know what full ISUP transparency gives you and why folks believe it is a complete requirement. What are the effects, inaccurate billing, poor feature infelction?4th June 2002 at 16:44 #21183GordGuest
Does anyone know how to convert a 7960 MGCP to Skinny?5th June 2002 at 09:28 #21184AlexanderGuest
I did it once, but don’t remember exact procedure. In the article I’ve used to do it described how to upgrade from Skinny (SCCP) to SIP, not MGCP-Skinny, but the procedure is similar:
Upgrading the Cisco SIP IP Phone Firmware
Dual Booting from SCCP or MGCP to Release 3.0