VoIP Gets a New P2P Routing Protocol (DUNDi)

bkw.org writes "Today Digium released DUNDi which can be used with the Asterisk Open Source PBX for p2p call routing. Digum has also released a whitepaper (pdf) on DUNDi so others can implement this new technology into their products and give VoIP a push into the mainstream." Voxilla also has a story.

Comments

P2P VOIP

[stanmann]

Great, Now I can get telemarketers trying to enlarge my bank software in nigeria, and viruses on my VOIP line and in my e-mail

Another great move

[salemnic]

* is awesome, and for anyone who hasn't given it a shot, I definitely recommend it. Digium even sells some FXS/FXO cards if you want it to replace your traditional in-house system.

Something neat for every geek!

s.

Is this a...

[davidsyes]

--crock
--crocodile
--crock o' dial
--dun deal

OK, cold & dumb...lame attempts at humor

Before everyone says...

[einhverfr]

What do we need this for? Don't we have enough protocols?

This is actually pretty cool from a distributed PBX perspective. I am not sure I would want to use it over the internet with untrusted PBX's but it would be pretty useful inside a large corporate structure.

The New Baby Bells

[I Hate Spam Alot]

Wonder if this will sping up into localized phone services like wifi and cheap fiber based internet did for small towns. The whole town chips in and gets to kick out the big corp. and run the local service as the citizan see fit. Plus it will be cool to run in my house!

Cool

[LittLe3Lue]

Maybe soon I'll be able to call l33tMovieRipper over xyz P2Pclient and ask for The Matrix in person.

>

Language trivia of the day

[Anonymous Coward]

dundi means chubby in Hungarian.

Here come the security problems.

[IGnatius T Foobar]

Remember when closing off your email servers from open relaying was a skill that not everyone had a grasp of? Here we go again. While I do agree that VoIP inevitably requires more advanced routing, it is my fear that this will be abused for a long time until admins become skilled in the art of preventing unwanted forwards. In the meantime, you'll have...

• Skript kiddiez abusing it to go VoIP-to-landline on someone else's nickel (oh yes, the days of phreaking are coming back)
• Bulk dialers. You thought telemarketers were bad? Wait until the spammers get a hold of free calls to your home!

I fear for the future... :)

Re:Well this doesn't seem very good

[fugginsuds]

I don't think p2p is the "worst possible way". Underwater with a tin can and string is a pretty bad way to start a conversation.

encryption? wiretap?

[r00t]

P2P changes things. The FBI might not like this.

Number assignment is an issue too, unless you
can dial a "number" that looks like an email
address or a URL.

Uninformative intro again...

[The Famous Brett Wat]

DUNDi stands for "Distributed Universal Number Discovery". It is a peer to peer system for locating Internet gateways to telephony services. Unlike traditional centralized services (such as ENUM), DUNDi is fully distributed, with no central authority. DUNDi is not a Voice over IP signalling or media protocol: it publishes routes which are in turn accessed via industry standard protocols such as IAX, SIP and H.323.

The above information is taken (with minor edits) from the dundi.com website. It's the sort of information that would have been useful in the executive summary, IMHO.

Re:Well this doesn't seem very good

[MacJedi]

Your history lesson is sound for P2P file sharing. Economic pressures can result in some nice innovation, eh?

However, as far as telephony goes, P2P makes sense as well. If you want to use the client-server model, you need, well, a server. Netmeeting for example requires this-- you connect to a server which handles directory info. A P2P VOIP network could decentralize this.

Where does the trust build from?

[CharlieHedlin]

It seems to me that the enum standard should work, and that this requires too much trust building.
In the case of a corporate PBX, couldn't enum be used with distributed redundant DNS servers?

I use Asterisk, it is great, and I like the idea of DUNDI, but by the time you get a web of trust built, why not just use enum?

Re:Where does the trust build from?

[Gaewyn L Knight]

Easy answer... ENUM requires that you trust a central body in control of the use of your records. Havn't we all learned from DNS and Verisign?

Re:Where does the trust build from?

[Scott Laird]

Well, look at things this way: ENUM has been around for *years*, yet there's still no official ENUM tree. There's no way for VoIP carriers like Vonage to publish SIP addresses for the PSTN numbers that they service.

Why? Because of political squabbling by telcos, verisign, and the like. Whoever controls the ENUM tree will be able to control the future of telecommunications. That means money and power, and that's why there's no progress occuring--all of the usual players are jockeying for position, and will be for years.

Since ENUM is really just a DNS tree, that hasn't stopped people from producing their own ENUM trees (e164.org, etc), but there's nothing particularly official about any of them. They're all interim solutions, and none of them are big enough to be able to make a difference on their own.

There are a couple differences with DUNDi. First, it's *designed* to be decentralized, without a single point of control (or toll collection). There's an open-source implementation right out of the gate. It at least pays lip service to spam and telemarketing issues. As long as you sign the agreement, it *should* be possible for anyone to participate. And, it already has several mid-sized providers involved.

In short, right out of the gate, DUNDi is already ahead of ENUM, because it's already usable, while ENUM still doesn't have any way to publish numbers in the "official" e164.arpa tree. DUNDi doesn't have room for Verisign-style toll collection, while the official ENUM tree almost requires it.

We'll see how it goes. If Vonage joins up, then DUNDi has probably won and ENUM will end up being irrelevant, because the network effect will strongly favor DUNDi.

Re:Skype

[DarthBart]

No. Skype is still as proprietary and closed source as ever.

Re:Well this doesn't seem very good

[ZX81]

This new protocol is not for peer to peer voice communication. It's a method like a phone book. So you ask your peers if they know how to contact person X, your peers contact their peers and so on until they find the address of a person (for example the IAX/SIP/OpenH.323 address). Then the address is passed back though the peers to the original person doing the lookup, and is cached along the way.

You then make your telephone call using VOIP. The IAX (Inter Asterisk Exchange) protocol is amazing at getting though NAT'd connections etc. It can even trunk lots of calls together into one packet.

So in a nutshell, this is like a p2p enum.

Skype is now Obsolete

[Bob_Robertson]

And they did it to themselves.

1: Skype uses proprietary protocols that are incompatible with any other service.

2: Skype refuses to publish their interface profiles, so no one else can make software that is compatible.

3: Skype has stated in no uncertain terms, over and over on their web page and FAQ, that they will never publish their API, never open their source. Period.

Yet Skype doesn't sell their software. They maintain the full burden of development and testing, and try to pay for it by selling the service of POTS calls made through their software.

Oh well, once the idea is in the wild, someone will "open source" it. Maybe someone will adapt the Speak Freely user interface to a P2P transport layer, maybe utilizing the Gnutella network itself? Hey, this is sounding like a good idea....

Bob-

Skype and Standards and Alternatives

[billstewart]

Skype prefers to do things that are "better" than the current standards, and does them in a proprietary way to protect their potential profits, and doesn't document their protocols because they're not interested in having random people develop software that interconnects with them. (Make whatever judgements you'd like about that... :-) However, unlike many vendors who take that approach, they've at least done a decent job of it.

• They're using audio codecs from Global Ip Sound, who make codecs that are more tolerant of high packet loss than most of the low-bit-rate public-standard codecs, and also better-than-telco-quality higher-bandwidth codecs. It looks like Global IP is working on getting some of their codecs into the standards track.
• They view NAT traversal as a critical design element, because NAT's become extremely widespread (in spite of being evil breakage of the end-to-end paradigm), so they've done more than the SIP standards do to simplify that. (SIP came from Internet people, so it was far, far better than the H.323 stuff that came from the ISDN crowd, and it's easy to set up firewalls for SIP transparency, but NAT traversal takes extra work.)
• They view security as a critical requirement, so they've got modern crypto algorithms like AES in there, and from a performance standpoint it's a really big win to encrypt the data packets rather than using IPSEC tunnels, because VOIP data is inherently small compared to the headers. Unfortunately, because of their attitudes about proprietariness and no public documentation, it's not possible for the crypto community to examine their protocols or code, and most crypto mistakes these days are made in protocol implementations, not in the fundamental algorithms, so even though they use AES and Diffie-Hellmann and long enough keys, that doesn't mean they're not totally hosed.
• P2P is fun, and can scale well by taking load off the central server, and the Skype folks don't want to run a huge central server. This has some conveniences for their design (supernodes for NAT traversal assistance, etc.), and creates some interesting security tradeoffs (no central point of attack, but widely distributed local attack points) which are unfortunately covered up by the lack of protocol documentation.