Blog Post

Telephone as a platform, deconstructed

Last week I moderated a session that was supposed to discuss “how rise of IP (Internet Protocol) has turned telephone into the new platform”. My favorite VoIP guru, Aswath Rao, heard the MP3, and came-up with things which were not addressed at the panel, where I screwed up, and how the guys on the panel sent mixed messages. Here is what Aswath has to say:

The panel, mostly turned out to be mud fest between AT&T and Vonage. Given that you can hear the full session and many have blogged the event and my tardiness, I am not going to summarize the discussion. But I would like to tell you my reaction as I was hearing the session; it is a good opportunity to get on my soapbox.

  • Eslambolchi (AT&T) says that in PSTN, phones were dumb and that switch did all the work. But that has all changed in VoIP. (No, don’t ask him how this is so in CallVantage which uses MGCP.) That is why they are not calling it VoIP, but S(ervices)oIP. If the intelligence has really moved (and allowed to move) to the edges, then won’t the services be realized there. If so, then what is the role of the service providers? The forum didn’t address this point.
  • McCue (Tellme) says contradictory things: VoIP and VXML is a winning combination (according to CNET, he has been waiting for VoIP to take off) and there are infinite (2B was it?) PSTN phones that needs to be served. But nobody calls him on it. He chides AT&T and Vonage that they are thinking like phone companies (it is a put down if you didn’t know). He tells them that they should allow others to use the billing system so that others can offer “exciting new services” for a monthly business. In other words, everybody wants to be a phone company. In any event, what is an example of an exciting new service? It is Voice mail. Nobody tells him that voice mail could be realized by the customers for no additional cost . It looked like everybody is satisfied if there was an ARPU – the curse that is making VoIP look like POTS with a sex appeal. The same technology segment comes up with GMailFS that treats an email account as a file system, but tells us that we need something special to store voice messages and nobody raises an eyebrow.
  • From the floor Marc Cantor asks exasperatingly when will we have interoperability between different service providers and the rest of the audience agrees with a thunderous applause. Instead he should have said to the panel that we don’t need them because we can realize all the benefits of VoIP by ourselves without contributing to anybody’s APRU.
  • Nobody asks Om about the glaring absence of terminal vendors so that we can ask them why they are marketing their wares to the service providers rather than directly to us and when will they improve on the user interface, but maintain 12+1 button user interface.
    In short the forum did not instigate a change to the business paradigm; indeed it reinforced the false promise of ARPU business model.

8 Responses to “Telephone as a platform, deconstructed”

  1. The residential gateway in my comment is the same as the ATA in yours. Actually, many technical documents show that (the ATA/ residential gateway call agent being controlled by the service provider) as one of the possible network architectures, which makes it a possibility that AT&T’s choice could be driven by technical matters as well, not just by the urge to *control* service delivery. Need to dig into stuff to figure this one out.

    Another discrepancy about VoIP that I am not able to reconcile is that on one hand, people talk of VoIP being about new services delivered on top of the network (this post itself is titled “Telephone as a platform”). However, at the same time, lot of people predict intense price competition (it is even being played out right now between AT&T and Vonage) among VoIP players. So I am eager to hear thoughts on whether VoIP will be just commodity voice, with consumers buying from the cheapest provider, or will consumers be willing to pay a premium for “new services”? Or will the latter be just limited to a niche, just as Apple is, while the mass market is dominated by the cheapest provider(s)?

  2. Level3 may very well shine in interconnecting VoIP and PSTN; but I do not see their MOAT. Once the interconnect business reaches a critical mass, any of the incumbent players can deploy the gateways and utilize their existing network capacity.

  3. the Media Gateway and the interface to connect all voip to PSTN is my main reason for touting Level3 so much! What folks do NOT recognize is that Level3 is building a MOAT around handing off all the “”ON NET”” Voip traffic back to the PSTN!! I for one had thought Level3 would be the “”ON RAMP”” to Voip but it turns out anyone can get traffic ON NET but its getting the VOIP off to the PSTN where Level3 is going to shine!!!


  4. You are right that MGCP or equivalent is used to control PSTN gateways. At the same time, an ATA can be viewed as a kind of PSTN gateway. That means MGCP can be used to control the ATA. Many variants of this protocol (like ’Äúskinny’Äù from Cisco and NCS from Cable Labs) are being used for this purpose. It is my understanding that AT&T uses MGCP in this fashion. It is very much the case (in AT&T architecture) that MGCP is a component of VoIP interface. The ATA that AT&T ships to its customers has one and only master, AT&T Call Agent. I am suggesting that it is a problem.

    For all practical purposes, this allows for the service provider to recreate the same business model as in PSTN ’Äì the customer can derive the services only from the provider, no matter how intelligent the terminal has become (as Eslambolchi stated in the forum).

    Regarding PSTN interconnect, I am suggesting two points. First and foremost, we should develop VoIP without any regard to PSTN interconnection. This is because with VoIP many new capabilities can be made available to the end-user. Second, the architecture should allow for the customer to receive interconnection service from multiple providers as it is in PSTN. Think of ’Äúdial-around’Äù. The current arrangement goes back to the monopoly days of Bell system.

    I have a simple way to address the reason you are suggesting to eliminate the need for ’Äúan AT&T’Äù, but I am not ready to share it yet. Some day soon, I hope.

  5. I am sure I am missing something since my telco experience is dwarfed by yours, but my understanding is that calls between SIP endpoints and PSTN phones have to necessarily traverse through a media gateway when they move from the IP network to the PSTN network. Thus, a SIP-PSTN call requires MGCP/MEGACO functionality (to coordinate the call setup and media path) at the “other” edge of the IP network which connects to the PSTN.

    Or maybe, you are talking of an architecture where SIP phones “plug” into a residential gateway which the service provider, AT&T in this case, controls using MGCP. I am not so upto date on this one, but it seems the advantage of this approach is the ability to control/configure multiple SIP endpoints in a residence through a single interface. Implementing “intelligent services” (through call processing logic) for a user with multiple SIP endpoints — a key benefit of SIP architecture — is easier when a single entity is aware of all such endpoints. Compare that to the cumbersome procedure of “planting” call logic on all SIP endpoints that I use over the course of the day (IM, Voice, Video, Mobile etc).

    In either case, MGCP is not a component of the VoIP interface (endpoint) but sits somewhere in the network and using a SIP phone does not mean that you no longer need MGCP to connect with PSTN.

    Regarding the two stage dialing, sure that is an option, much like people have now for long distance dialing – once I sign up with SBC for dialtone, I could either purchase a calling card, or subscribe to SBC (or ATT) for a direct-dial long distance facility. Similarily in the VoIP domain, once I have my broadband connection from Comcast, I could buy a “PSTN interconnect” calling card (as you suggest) , or pay AT&T a monthly fees for the same. As long as some people prefer to do the latter, AT&T will be around.

    And another reason why I feel that it will be difficult to eliminate the service providers is the need for 99.999% uptime – for times when my SIP node is down or rebooting (very likely if I use a MS box to run my SIP proxy or endpoint software), I need intelligence in the middle of the network to appropriately handle (redirect to voicemail or cellphone) SIP calls headed my way. And as long as the network is not totally dumb, there will always be a need for an AT&T, albeit only a shadow of the giant it has been in the past.

    My $.02.

  6. I don’t get it why MGCP is needed on the VoIP interface to interconnect to PSTN? FWD, SIPPhone, Vonage et al. are able to do it.

    If interconnection to PSTN is the only reason to subscribe to a service provider, then use them just for that, like a calling card call. It is two stage dialing, but the user interface takes care of the second stage.

  7. If the *entire* installed base of POTS phones were to morph overnight into VoIP phones, AT&T wouldn’t need MGCP – SIP, which is more of a P2P protocol, would suffice to connect two terminals. Actually, in that scenario, customers would no longer need AT&T (or any other service provider).

    But until that happens and people continue to use POTS phones, there will be a need for providers like AT&T, which in turn will have to depend on MGCP to serve its customers.