If you look at SIP today, it is the most widely used protocol for VoIP when it comes to developing telephony services. In a way, it overthrew H.323 in most areas. On the other hand, in the past several weeks I’ve been bumping a bit more into XMPP. To some it may seem as if XMPP is set to overthrow SIP as the VoIP signaling king.
XMPP is Becoming a Swiss Army Knife
I have already tried comparing XMPP with SIMPLE for presence, but it seems like people have already commented there about the advantages of XMPP, and they did out of context of presence service alone.
I was skeptic about the use of XMPP out of the pure presence related arena, but things might be changing. I’d like to point out three changes.
1. Google Video Chat
It is no secret that Google is using XMPP for their own GTalk service. They are doing that with the Jingle addition to XMPP, which provides VoIP calling on top of XMPP (the main service SIP is used for today).
To those who haven’t yet noticed, Google have launched their video chat service in Gmail. While this has nothing to do with XMPP, it does say a thing or two about the potential of Google’s service to grow fast now that it supports video chat.
2. XMPP for Everything
I bumped into this idea of using XMPP to bridge with telnet. By the looks of it, people are going crazy with ideas of what to do with XMPP. This is a bit similar to what happened with SIP a few years ago.
3. XMPP vs. SIP
If you think about it, SIP is used for VoIP while XMPP is used for presence. Period.
Google is currently the only user of XMPP for VoIP, but that might change depending on how successful their service becomes.
While SIP is a session based “thing”, XMPP is a presence based “thing”. Each is being adopted for a lot of additional uses – some more suitable than others.
From a technical point of view, the question is: does it make sense today to use a communication protocol that focuses around VoIP or a communication protocol that focuses around presence (especially when thinking about Unified Communications).
From a standards point of view, SIP is gaining ground with operators and service providers through IMS, while XMPP is gaining ground as the presence solution for Unified Communications.
Is that going to change? Is XMPP set to overthrow SIP through its larger deployment base in presence solutions?
Tsahi my friend,
A good question.
I did a gap analysis for GSMA across all IM/Presence protocols and, personally, I preferred XMPP and couldn’t see why it wasn’t adopted. As you know, XMPP can also be managed via SIP session, if need be.
I would say that the whole issue of presence is much more important for the future of unified/converged comms. Operators should quit aligning all such ventures with IMS availability and get on and release XMPP phones with presence-enabled address books driven by a central presence server with OPEN API for subscribers to subscribe to any presence info they like within their data plan, including Twitter.
This will open up a new wave of innovation and interest in data services, as we shall soon see with the INQ1 phone on 3UK network.
All the best and keep blogging!
Tsahi,
Good article. We have been discussing about this on your blog in the past and I believe XMPP has the advantage because I thing the protocol and architecture are much cleaner than in SIP.
A good question is on how suitable XMPP is for setting up media sessions. Would XMPP be good enough for an operator to run its business on?
A nice feature of SIP is that its transaction based. If you send out a presence update you can be sure that it is received. For 99% of the presence usage this might be unnecessary. But if you are building applications where it is critical that a presence update is delivered correctly XMPP does not help you out-of-the-box if I understand things correctly. I could think about some application scenario’s that would need this kind of transactional function.
Since I am a pragmatical person I believe for these exceptions you can build something special on the application level in XMPP to achieve the same thing.
Thanks,
Eelco
Apple iChat also uses XMPP.
Interesting to see this coming from a Radvision guy – we use the Radvision SIP stack at my place.
On the other hand, I’ve been exploring XMPP for some time. I agree that in the field of UC it looks to have an edge. Where I think the other edge is: will it’s hinted at in your post. More and more interesting uses of XMPP are cropping up, and being deployed on the internet. So I see the other edge being a more generic way to add value beyond traditional UC (if traditional is a term that can be applied to UC.
BTW, I’m no IMS guy, but my reading seemed to indicate that IMS does not use SIP for authentication/authorization, but uses DIAMETER. SIP does not have the best security profile going, XMPP would seem to be ahead here.
But anyway, your distinction of Internet versus voice carrier is still spot on. So another way to look at it is – will the 3GPP view of mobile/end-point-independent communications be the dominant force shaping that world, or will it be the Internet view of the same?
I know which way I would bet.
@paul – I agree that presence on handsets is needed. For the purpose of the discussion, I wouldn’t care less which technology is used as long as there’s a solution. That said, operators doesn’t seem too keen in having that yet, and there are also technical issues to overcome for large deployments.
@Eelco – thanks for the feedback. I guess the thing is that VoIP protocols can do everything if you want. All features in H.323 can be supported by SIP and vice versa, and I guess the same can be said for XMPP. Today, SIP is what is used for VoIP and I really don’t see it changing anytime soon. XMPP might replace it, but not in the near future.
@tony – thanks! I didn’t know that bit of information. Do they use it only for presence or for the VoIP part as well?
@thomas – I have been “looking around” lately, trying to make sense of what happens in the VoIP industry where standards are concerned. I believe that for a VoIP centric service that needs presence, SIMPLE can be a better solution that XMPP (being more tightly coupled with SIP). For an IM client that does a bit of VoIP, XMPP is probably preferable.
As for DIAMETER/SIP – DIAMETER is what you use for that, while SIP is used for almost all the rest – 3GPP wisely selected DIAMETER for that (this is what it does, and almost only that). As for security in SIP – IPSec in IMS and TLS elsewhere takes care of that part.
Oh – and your question of 3GPP versus Internet guys is an issue I’ve been pondering a lot. I like Free as the next guy, but I have a feeling that 3GPP will prevail in the long run. And that’s my own opinion, unrelated to the fact I’m a “Radvision guy”
I’m confused – when you talk about “an IM client that does a bit of VoIP”, you’re essentially describing the current usage of most mobile handsets – people use them for SMS a lot, and voice comparatively little.
And given that the killer addition to VoIP – the bit that makes VoIP surpass traditional telephony – is presence, and XMPP appears particularly strong in that area, my inclination is that XMPP might even be better suited to VoIP.
Finally, I would note that all the recent presence-based applications for mobiles appear to be using XMPP as a base – Nokia’s Chat application, Jaiku, etc.
Of course, I’m undoubtedly biased, and my view of the world is coloured by my surroundings.
A couple of fact checking exercises, BTW – Google’s video chat is being worked on by Collabora, and does indeed run over XMPP. (Albeit not Jingle, per-se – Jingle has moved on, and Google haven’t.) Collabora also produced an independent implementation of Google’s Voice protocol, which formed the basis for Jingle.
I suspect Jingle’s “attitude” to VoIP might win out, in fact – it’s much simpler than SIP, with the additional, more complex, telephony bits moved to extensions, meaning it’s faster to get up and running.
@Dave – SMS is far from IM with a different approach to communication. You can see what I’ve written about it in the past.
Today, at least for me, most of the communication is VoIP based. I use presence mainly to IM, and less to VoIP (in terms of how I “upgrade” my communication channel), though I have that option available to me.
For my daily work life, VoIP and IM/Presence are separate in my perception and in the way I work. This maps XMPP to my presence life and VoIP to my SIP life.
That said, I VoIP with Cisco (proprietary protocol) and IM with OCS (SIP based).
As for XMPP being simple – I suggest you check out what Ran from the sister blog here wrote earlier today about simplicity and complexity of VoIP protocols – if XMPP will win in the end, it will probably end up being as complex and as muddled as H.323 and SIP are these days.
Oh – and thanks a lot for the valuable information on Google’s use of XMPP! It’s much appreciated.
Hello,
Today you can do VoIP call with a Google Talk user if you use Empathy IM client. ( http://live.gnome.org/Empathy )
Empathy uses the telepathy specification ( http://telepathy.freedesktop.org/wiki/ ), so it has support for XMMP, jingle, msn , aim… and SIP
This especification is used by Nokia for the maemo platform (N810).
But Where is H.325? Is there room for it?
@Javier – H.325, which is also known as AMS for the time being (there’s a discussion on a name change these days) is not a working protocol yet – it’s a protocol in the making.
It will take several years until it becomes a standard that can be implemented. The way it is designed is by looking at what’s broken with today’s standards and trying to define something that will be more robust.
It looks at a post-IMS world. So there’s room to it, but not in the time being.
Pingback: On SIP Complexity and XMPP Cleanliness - VoIP Survivor
@Tsahi Levent-Levi – yes but, is really necessary? Not is possible to make all the things that H325 said with XMMP for example? Maybe I’m wrong but I can’t find anything in the H325 papers that I can not make today with XMMP (or tomorrow with a bit of work).
What I mean is that H325 would be a waste of time if we can improve actual protocols for do the same things.
What do you think?
@Javier – that is a very good question which I am afraid I don’t have a good answer for.
The point is that you can do anything today with virtually any IP based communication protocol. The difference between these protocols is in their base architecture and what they have evolved into.
H.235 comes as a protocol with no legacy attached to it and tries to see what is the best architecture to fit the needs we have today (and tomorrow, if that can even be forecasted in a reasonable way).
XMPP has its legacy, and as all protocols go – it probably has its own set of architectural issues. Take into account that improvements on protocols can always result cumbersome results.
That said, H.325 is still to be defined and its future is questionable and depends a lot on the level of commitments companies in the industry will give it.