The best days of VoIP are yet to come

March 7, 2005 - 11:44am — bmann

Erin Dalzell sent me a link to a recent Cringely article ("The Best days of Voice-over-IP Telephone Service May Already Have Passed"). Cringely seems to think that VoIP is doomed because it will just be sneakily blocked (or rather, not tagged with QoS.

I think Cringely gives the telcos and big network operators too much credit. This would be non-trivial to implement across entire networks.

Aside from the difficulty or not of tagging all network traffic is the encrypting/tunneling option: much like P2P traffic that will likely start to travel encrypted over common ports (e.g. pretend it's regular HTTP traffic or VPN traffic or...), VoIP has the same option.

If the network doesn't know you're running VoIP, it can't block it.

Suffice it to say, I'm not too worried about the future of VoIP. We haven't seen anything truly interesting yet (well, Skype comes close, but it's closed, proprietary nature makes me not want to like it), and there is lots of room for innovation yet.

VoIP

Comments

March 22, 2005 - 1:35am — Aunty Spam (not verified)

Duane wrote: "Yeah, I d

Duane wrote:

"Yeah, I don't agree with that article at all.. Recently someone tried to block VoIP traffic, and they were slapped with a $15,000 fine.. It's unlikely many more companies are going to try that for a while."

Actually, despite Madison River being b*tch slapped by the FCC for $15k, Vonage is claiming just this week that they are being blocked by two more carriers.

I'd expect it to continue happening until a court really puts the hurt on providers who do this (how painful do you think a paltry $15,000 is to a provider who is losing revenue to one of those upstart newfangled VoIP thingamajigs? - and by that time it could well be that we'll have additional regulations in place such that the VoIP picture, from a regulatory perspective, will have already radically changed.

Kissy kissy,

Aunty Spam

March 19, 2005 - 10:02am — rogot (not verified)

Cringely's wrong..to some degree

ISPs are not going to waste precious routing cycles to tag packets as ours/theirs and apply CoS markings at ingress. It simply takes too much horsepower at the provider edge. They will simply choose whether or not to honor the TOS markings that were applied to the packets as they were created. Do not expect ISPs to create TOS tags on a customers behalf, but rather to accpt and translate your TOS markings to their internal TOS scheme. As ISPs get MPLS (and TE) into their backbones and finalize their own internal traffic prioritization schemes, expect to see customer facing QoS packages as the upsell from your ISP. You can buy phone service from Vonage, but you should also pay for MY QoS package, or you can buy VoIP service direct from [ISP] and get QoS for free....

March 19, 2005 - 8:32am — Anonymous (not verified)

Anyone who has worked with ne

Anyone who has worked with networks realizes that this is indeed possible, if not easy. Most telco networks are really ATM-based, which already supports QoS. They would also know that it is not feasible to tunnel a connectionless protocol like UDP, which carries most VOIP calls, over a VPN, which uses connection-oriented TCP, because you lose the benefit(the speed and efficiency) of the connectionless protocol. The only way to stop it would be a law stating that they must give equal preference to all providers.

March 9, 2005 - 6:36pm — gfox@muddylaces.ca

There is lots happening in the SIP world

or so it seems anyway. There are plenty of existing apps that have some interesting functionality, like a P2P file sharing app based on SIP.

Currently I am working for NewHeights on their SIP based product. Very cool stuff when it gets working, but the way I see it, SIP has some huge hurdles (that can't get solved quickly) before it can be ready for mainstream consumers, and to replace local dialtone. Until it can do that, I have no reason to use VOIP at home.

Another interesting VOIP web-app is the switchboard.

The way I see it, the best days of VOIP are still to come.

March 10, 2005 - 9:53am — Duane

Duane

There's still a ton of work to do in the P2P front.. The real trick is to be able to do it without any supernodes, and have a full self-configuring network. That's essentially the holy grail of P2P, and something MIT has been working on for a while.. I've looked at a few different methodologies, including CHORD, for P2P.. The main problem is that all the schemes fail to take into account real world problems.. almost all VOIP is over UDP, which means you have to do keep alives periodically to make sure your NAT doesn't close down your connection on you. We calculated the pure bandwidth required for *just keep alives* on a large self-configured P2P network one time, and the numbers were quite staggering.. If you don't mind supernodes, it works today.. but nobody really wants to be a supernode unless they get something out of it, so companies that release P2P software generally put their own supernodes out there, and hope for the best.

SIP is pretty wordy in my opinion, but it's really a no brainer for anyone on broadband..I mean, you can establish a SIP session in under 2k of traffic, so I really don't see SIP as any roadblock to getting VOIP into houses. There are a few real problems with SIP that just come out of the real world. For example, I have experience doing video using SIP.. On paper, it's not that hard to do, but in the real world, once you put video and audio in your SDP, often when the proxies tack on a few VIAs, you're over the link MTU, and some routers toss your packets.. So while trying to add something cool, i.e. video, you're inadvertently breaking the session establishment. The solution to this (at least on paper), is to establish the session over TCP.. That actually works quite well, but unfortunatley most of the deployed equipment doesn't support TCP. I'll actually be attending SIPit this year in Banff to help with interoperability Sip tests.

I think by the end of the year, you'll see many SIP endpoints, including hard phones.. I was talking to someone at Cisco a while ago who said that they can't believe how many hard voip phones they sell these days. So, I really believe the next 12 - 18 months will be VOIP's time.

March 8, 2005 - 4:25am — Duane

VoIP packets

Yeah, I don't agree with that article at all.. Recently someone tried to block VoIP traffic, and they were slapped with a $15,000 fine.. It's unlikely many more companies are going to try that for a while.

QoS is almost here for VoIP. Many people have argued that unless the network supports it all the way from end to end, that implementing QoS doesn't make sense.. But it's really only the first hop or two that's important -- after that most people have hit the open internet, and are travelling along decent equipment / lines. Even if the QoS is at your OS stack level (for soft phones), it's still a major improvement, since VoIP traffic will get priority over all other traffic (you can even differentiate amongst streams within a VoIP call, i.e. prioritize audio over video).

Encryption is also almost a reality for VoIP. The SIP traffic can already go over TLS, and secure the establishment of the session. Unfortunately, encrypting media is a little trickier. The standard most people look to these days is SRTP, and there's a good open-source library from the boys at Cisco called libSRTP.

There's still a lot left for innovation in the VoIP front.. In terms of hardware you're going to see the slow consolidation of gadgets into some kind of blend of SIP based phone/organizer. There's been a lot of research in self configuring networks. Intel just announced a new wireless 802.11 protocol that supports a self configured wireless network. Combine this with some smart auto-configuring sip proxies (sip-config is also a new standard nearly completed, which allows users to automatically locate proxies and utilize them), and you're going to see some cool apps in the future.

March 8, 2005 - 10:59pm — bmann