I am a huge fan of Martin Geddes, but he and I disagree fundamentally on one key part of what he is now calling “hypervoice”.
To back up a bit, Martin has always been one of the “big thinkers” in realm of VoIP and telephony/telecom. Way back in mid-2000s when a number of us all started writing about VoIP, Martin’s Telepocalypse blog was brilliant. He was always thinking about the “big picture” and drawing connections where they were not already apparent. His work with “Telco 2.0” was excellent and it was no surprise when he went to work for BT looking at their strategy. Now that he is back out on his own as a consultant, I’m a subscriber to his “Future of Communications” email newsletter (subscribe on the sidebar to his site) and enjoy reading his frequent issues.
Recently he gave a closing keynote presentation at the Metaswitch Forum titled “A presentation about Hypervoice” that is available via Slideshare or PDF.
The presentation itself is very well done. In typical Martin style it nicely lays out the history of both telecom and the web and brings them together to talk about what comes next.
I actually agree with almost all of what Martin writes. Much of what he talks about as “hypervoice” I see already happening in so many ways.
But here is where we fundamentally disagree… this slide early on:
That includes the text:
“However, the Internet cannot and never will carry society’s real-time communications needs. It is fundamentally unsuited to the job.”
Martin’s argument, which he has made multiple times before, including in a comment he wrote in response to my post about how WebRTC will disrupt real-time communications, is that the Internet as it exists today cannot provide the level of service that is truly needed for real-time communications. He believes we need to have different classes of service on the Internet and separate “flows” of communications. He comes back to this point later in his “Hypervoice” slide deck:
This is where he and I part ways. As I said in my own response to Martin’s comment to my earlier post:
Martin, yes, I’ve read your newsletters on this point and while I understand the concern I’m not ready to say that the plain old Internet can’t deal with the contention. Back in the early 2000’s I was the product manager for Mitel’s “remote teleworker” product and there was great concern from the traditional telecom folks within Mitel about this idea that we were going to put an IP phone out at some random point on the Internet where there was no QoS or anything. In fact, some folks wanted us to say that it had “cell-phone voice quality” so that we wouldn’t set high expectations about voice quality. The reality was that through appropriate codecs, jitter buffers and other technologies the connections almost always worked and almost always had outstanding quality (usually FAR better than cellphones).
The other reality is that we’ve seen OTT providers like Skype and others providing excellent services that work the vast majority of the time. We’re seeing new and improved codecs coming into the market. We’re seeing new traffic shaping technologies. The list goes on…
If the (brief) history of the Internet has shown us anything, it is that the Internet’s capacity to adapt and change is boundless. We’ll see what happens in the time ahead.
And no, I haven’t written off the telcos as having a role in real-time comms. I just don’t know that the “role” they may have will necessarily be the one they would like to have! 😉
I believe fundamentally that the “open” Internet can and will adapt to the needs of carrying real-time communications. I would argue that it already has in so many ways… and it will change even more as we continue to move more and more real-time comms onto the Internet, particularly with WebRTC and other emerging technology.
And yes, you might expect me to say this as a passionate advocate for an open Internet, but I firmly believe this:
We do NOT need separate layers of the Internet based on class of service.
That, to me, is a dangerous path. I want to continue to see an Internet where all nodes are treated equally … and where real-time communications can work for all.
Martin and I will probably have to agree to disagree on this. It’s doubtful he can convince me nor I can convince him.
What do you think? Do we need different layers of the Internet? Or can the Internet adapt without that? Leave a comment here… or join in to today’s VUC call and comment there.
If you found this post interesting or useful, please consider either:
- following me on Twitter;
- adding me to a circle on Google+;
- subscribing to my email newsletter; or
- subscribing to the RSS feed
See my article http://us1.campaign-archive1.com/?u=f105fd56904428bca9da44a82&id=0e316c1c15
To carry society’s real-time needs, “works most of the time” isn’t enough; we need a level of predictability of performance that the Internet is innately incapable of delivering. It’s about containing the outliers, not raising the averages. The Internet has too many outliers, and they cannot be resolved through overprovisioning. A multi-service world already exists — we bleed real-time traffic away from the rest into overlays — the question is how we want to go about it in a world where the diversity of needs of application flows keeps growing.
Traditional approaches to QoS are both philosophically and technically flawed. I had better get on with writing my book on why, and how to fix it…
I think the internet is already operating in different layers.
You can place a web server somewhere and see how slow it is to get its service running on the other side of the planet, and you can then pay to have a CDN to serve your files for it.
WebRTC will go through the same path: people will start deploying servers around the world that will reduce the time it takes to do NAT traversal, relay media or even run over virtual MCUs.
Thinking about it this way, why not pay to get a lower latency from the service providers as well? How different is it from buying an MPLS line as an enterprise or by CDN services as a “broadcaster”?
Martin,
Yes, I remember that article well as it was thought-provoking when I first read it.
I don’t disagree that there is a challenge as we move more and more real-time communications onto the Internet, but where we disagree, it seems, is about the solution. If we look back at the history of networking and standards development, there were numerous attempts to develop various systems that contained all the outliers and edge cases (ISO, anyone?).
They all generally failed. The best-effort approach of TCP/IP proved to be “good enough”. And my argument is that as TCP/IP and the Internet in general continues to evolve, more solutions will come into play to assist in enabling the RTC that we want.
An example is the work currently going on within the IETF around congestion control related to RTCWEB/WebRTC. The result of that work may in fact turn out to help address the flow/congestion issues brought about by WebRTC. Other efforts are going on within other parts of the Internet on other similar issues.
Dan
P.S. Your book on a new approach to QoS would definitely be worthwhile to have out there!
Does hypervoice require the reliability and QoS of the PSTN? I don’t think so – I’d argue the whole ecosystem has changed:
http://goo.gl/gJXuU
Hypervoice still requires a real-time audio stream if people are talking synchronously, however it is not necessary for that to go over the Internet. Any medium that actually works with the reliability the user chooses is OK. The Internet is a natural medium for uploading locally-recorded audio; after all, it’s designed to to FTP and telnet well .
“We do NOT need separate layers of the Internet based on class of service.” and “… and where real-time communications can work for all.”
These two statements are in conflict. You can’t have a single class of service, reliability for real-time services and low cost at the same time. Pick any two.
“You can’t have a single class of service, reliability for real-time services and low cost at the same time. Pick any two.”
I’ll go for all three: a single class of service, ‘good enough‘ reliability for real-time services and low cost.
Martin, a.k.a. Mr. Telepocalypse…. you know the Internet works well for far more than FTP and telnet!
Sure… the real-time audio stream could go across some private or separate network if the people in the call had access to the separate network and wanted to pay the costs, etc., involved with maintaining that network. We already do that today with some of the OTT mobile apps where you have chat/IM conversations over the mobile data network but when you want to go to voice the call goes out over the mobile PSTN. And I could see a mobile app that might check the available bandwidth and then initiate a call over the Internet if possible and fall back to the PSTN if it can’t.
But as the codecs get increasingly better on the IP side and as improved congestion control algorithms come into play, I question whether we’ll need the separate networks.
Yes, the Internet is fundamentally a “network of networks”, with many, many different networks operating at very different speeds, bandwidth capabilities, etc. You pay for the speed you want for the services you are offering. Buy a bigger, fatter pipe and you can have faster access to your web server… or can have more simultaneous calls.
However, all of those packets have the same “class of service”. There aren’t any “special lanes” on the highways where some packets get more priority than others. It’s just that in some areas you have 8-lane super-highways while in others you have 2-lane highways or even 1-lane roads. You need more speed you just get a bigger road. Or if you send a lot of packets to a certain destination, you look at how you get connections that are close in terms of network connectivity, i.e. you make a road that goes into the nearby town.
What Martin is advocating (or at least what I understand him to be advocating) is that the networks should have separate “classes of service” where some packets get more priority than others. Think of it as special lanes on the highway. If you can pay the additional fee, your packets can receive priority routing. So now you are not only paying for the bandwidth (the road), but you are also paying for access to the special lanes. And you are now at the mercy of whoever is controlling those special lanes in terms of whatever fees they want to charge (or increase) and in terms of potentially whatever content they want to allow over those lanes.
I don’t want an Internet with this kind of segregation and where content providers or service providers have to pay for special access to achieve reliability. I want that to all occur over the standard Internet.
I skimmed the above article and it has nothing to do with Hypervoice! LOL. So I’m rather confused about the title. To learn more what Hypervoice is, see this initial interview
http://blog.ecomm.ec/2012/12/Hypervoice-Interview.html
Lee
Lee, if you go back to look at Martin’s slides for his presentation at the Metaswitch Forum, which was the first place I heard him speaking about Hypervoice, he includes in his Hypervoice presentation commentary about how the existing Internet will not handle Hypervoice and how new networks need to be created. THAT was what I was reacting to in this post… and I was including it under “Hypervoice” because Martin was including that as part of his presentation.
Reading through your recent interview with Martin and Kelly, there was nothing about this aspect of Martin’s original presentation but instead it all focuses on the many aspects of Hypervoice with which I very much agree.
So this post may have nothing to do with how Hypervoice is now being described (which is good!), but it did have to do with how Hypervoice was being described when Martin first started talking about it.
Let me give you a different perspective, have a look at this video which talks a bit about the history of the Internet and priorities for streams and packets:
https://ripe65.ripe.net/archives/video/3/
The more interresting problems are with wireless networking and bufferbloat, routing packets at the core and the price of “transit” is still dropping so it is doing just fine thank you.