January 28, 2004

Redefinition of telecommunications and informational services

In a recent article, Konrad Trope, an attorney argues why we need to reevaluate the definitions for “informational” versus “telecommunications” services. He points out the inconsistency between the decision made by the Ninth Circuit Court of Appeals and a federal district court judge in Minnesota. I do not see the inconsistency.

This is excerpted from his article:

The federal courts are split on the issue. The Ninth Circuit Court of Appeals, on Oct. 6, 2003, classified cable operators as providing "telecommunications" services as defined in the Communications Act, when Internet access is involved. Thus, cable operators would be subject to FCC regulations that affect telecommunications companies.

Then on October 16, a federal district court judge in Minnesota ruled to the contrary and declared that Vonage, a VoIP service provider, is an information service provider. The court noted "Congress's mandate that the Internet remain unfettered by regulation."

I submit that there is no inconsistency. There is a difference between providing infrastructure service and an application service. Infrastructure service is capital intensive with a high barrier for entry. So there can be only a handful of such providers and it is usually very difficult for subscribers to change their service provider. This is not the case for applications. Usually the application is strictly end-to-end, requiring only network connectivity from the infrastructure provider. Even when a service provider is used, it is easy and undisruptive to change the service provider. This is the rationale for supporting regulation for infrastructure providers but keeping the application service providers’ regulation free. Earlier, I had argued that even telecom regulatory regime follows this line of thinking.

There are two regulatory news items that are of related interest. Recently, Telecom Regulatory Authority of India issued regulation on Quality Of Service for VOIP based International Long Distance Service, stipulating MOS level, end-to-end delay, limit on jitter and packet loss. Their focus is on the service providers who offer PSTN interconnectivity. I guess that is the rationale for their requirements on DTMF tones and data/fax modem support.

In a meeting yesterday California regulators and some of the VoIP service providers toned down their respective stance on regulating this industry. The regulators are willing to consider a lighter touch as long as 911 service and Universal Fund issues are addressed; the service providers are willing to go along as long as competitive environment is fostered. But Jon Arnold of Frost&Sullivan believes that facilities-based carriers have better chance of survival than “pure play” service providers because “aside from having deep pockets, they will counter with a more reliable offering that travels over their own managed networks rather than the "best efforts" public Internet.”

Posted by aswath at 01:36 PM | Comments (0)

January 26, 2004

Looking for Mr.Goodbar (unique features)

These days I feel I am in an echo chamber – I hear some set of statements about the new brave world of VoIP and the promised set of features. I want to present my analysis from a fresh perspective. But today I have nothing new to add. So I am just recording the statements.

Today’s story is from NetworkWorldFusion. There, Mark Kaish, BellSouth's vice president of data services says that regulators will treat packet voice on the same scale as PSTN and so he feels that reduced price will not be as attractive. “In the end, the features unique to IP voice will be the reason customers buy it, and at least initially that is business.” So what are some these features? The set quoted in the article is “Web interfaces to add and delete services, click-to-call, configure phones, see a list of who called, access calendaring and set the service to find users at different numbers.” Kevin Mitchell, an analyst at Infonetics says, “these are services impossible to deliver with traditional circuit-switched networks."

Posted by aswath at 04:11 PM | Comments (0)

January 25, 2004

Cloning of PSTN in VoIP Guise

Previously I have expressed my disappointment with the list of features touted by VoIP service providers. Today I read a story in US News that writes about new features facilated by VoIP technology. So I write about it once more.

The article quotes Gerry Campbell, a Time Warner Cable senior vice president. "The future growth of this business is through features, rather than price." He envisions “a wealth of applications developed by third parties, akin to software programs written for the pc.” This vision was stated many years back by Intelligent Network (yes the same IN that gave rise to the sarcastic “Stupid Network” moniker). But what is missing is the understanding why that vision more or less failed in PSTN. It might be true that "the telephone has been a closed shop, a closed arena." But that is not the whole story; there is a big technical point. The feature interaction problem is complex even when these multiple features are developed by a single organization and running on a single system. It is not clear how and why this problem can be solved simply because we are using IP.

Still, let us consider some of the exciting ideas the firms are touting:

  • Single number for life – ITU has a standard for it that can be used in PSTN. With nationwide roaming, wireless service providers are offering this feature. 800 number service suggests that the technology is available.
  • Central repository to collect all your voice and text messages – of course this has nothing to do with VoIP. A message server can collect the voice messages on the PSTN interface and then serve them on IP interface.
  • Call forwarding to multiple numbers – By the way this feature is patented and is offered as a service by some operators for their PSTN subscribers. When SIP describes “forking”, the possible numbers include PSTN numbers. If SIP Proxy Server can do this, why can’t PSTN Server (SCP of IN)?
  • Notification of voice mail messages via E-mail – again, why do we need VoIP for this. This is a capability of the voice mail server.
  • Using your tv and remote control to make changes to your calling system – once more we are confusing back-end functions with actual interface.

So I am still skeptical of revolutionary features we are going to see from VoIP technology. So what are the benefits of VoIP technology?

  • The startup cost for a new service provider is low (but that advantage disappears as it grows).
  • End-user can communicate directly without the intervention of a service provider or can change the service provider at will with little impact.

Posted by aswath at 11:09 AM | Comments (0)

January 23, 2004

Who provides three-way calling: Vonage or User?

Vonage advertises that they offer free three-way calling. Based on some web postings, I gather they use RTP proxy. So I thought they bridge the media flow in their network equipment. I came across a post that says: “Three-way calls: I took protocol traces of a 3-way call. I noticed that the ATA sets up 2 separate calls and does the voice path bridging in the ATA.” If this statement is indeed true, then can Vonage rightfully claim that they are offering free three-way calling?

Posted by aswath at 11:42 AM | Comments (1)

January 22, 2004

PSTN/VoIP Applications

Two recent news items appear not related on surface. But when I read them one after another, I get a different perspective of VoIP and PSTN. The first story indicates that Panama will treat VoIP and PSTN on an equal footing; the second story indicates that VoIP service providers will offer voice calling features that can’t be matched on traditional phones. This entry analyzes the feasibility of the last claim.

It is easy to see that governments would levy some form of “tax” on VoIP service providers in the guise of “technology agnosticism”. This has an air of fairness and the support of incumbents. I hasten to point out that I do not imply full regulatory regime, but I am commenting on the revenue sharing aspect alone. Alternatively, PSTN based service providers will be freed from their current revenue responsibilities. In any event, the tariffs for VoIP and PSTN will have parity. In my opinion, that is why many VoIP service providers would like to offer features that are not available to PSTN users. Are there such applications? The short answer is no. The one feature of VoIP that can not be replicated in PSTN is direct end-to-end session setup. But this implies there are no service providers. Almost all other features and services can be offered to PSTN users as well.

For example, the story starts with a feature offered by Rejection Hotline. Indeed this is currently available in PSTN. For VoicePulse to highlight this service indicates either a drought or somebody thought this is an easy way to draw a chuckle. If it is the latter, then they did not supply a more worthy example. Vonage’s example, Click to call is at least generally useful; but still doe not require VoIP. Vonage is planning to open up APIs, so that independent developers can build new applications. Interesting that “Intelligent Network” of PSTN is being revisited. If “thousand flowers are going to bloom” then how are we going to handle feature interaction?

Finally the article suggests that intelligence and bigger screens will be added to VoIP phones so that new features with better user interface can be offered. Of course PSTN tried that a few years back under the name of Analog Display Screen Interface (ADSI). The problem with it was the cost such terminals. With the current technology one can marry a low end PDA that sells for $25 or so to a cordless PSTN phone and realize an inexpensive ADSI phone. So it will be interesting to see feature war between PSTN and VoIP service providers.

Posted by aswath at 05:09 PM | Comments (1)

Is VoIP a square peg or a round peg?

In today’s NYT, the FCC Chairman Powell is quoted (requires subscription) as saying, “Don't shove the round Internet into a square regulatory hole.” According to Cambridge Dictionary Online, the common usage is “square peg in a round hole”. Did Mr. Powell give a deliberate hint? (At least you will conclude that I have a sense of humor, however labored it is.)

You engineers can figure it that given appropriate dimensions, a round peg fits a square hole better than a square peg in a round hole. (The proportion of fit is pi/4, rather than 2/pi). So probably what Mr. Powell is trying to tell us is that there may be some form of regulation and not the full regime applied to PSTN. Sometimes thinking like Oliver Stone is fun.

Posted by aswath at 01:26 PM | Comments (0)

January 20, 2004

It is not easy to offer QoS for VoIP calls

Many have lined up to offer VoIP-based services. This by itself is not surprising given that barrier to entry is almost nonexistent. Everybody has heard that the capital cost is very minimal. The rollout is not problematic. The situation is analogous to retail business. A brick and mortar retailer needs to open in many locations to have nationwide presence, whereas an online retailer can operate from a limited number of physical locations.

The same dynamics exists in PSTN vs. VoIP scenario. A single PSTN switch has a limited geographical reach. So it is more expensive for a PSTN service provider to support 1000 customers distributed over 10 clusters, rather than concentrated in a single area. On the other hand, a VoIP service provider does not have this problem. So it is much cheaper and easier to start rolling out a VoIP network. At least it looks that way.

A couple of days back Om Malik suggested in an entry in his blog that some people are complaining about quality of Vonage’s service. Some of the respondents identified quality of connection to be their concern. Also many established PSTN service providers who have announced their VoIP plans have indicated that QoS will be their main differentiator. This is easier said than done. First of all the service provider must remove the traffic from the third party access networks and move it to their managed IP network at the earliest. Of course architecturally Session Border Controller (SBC) will do the trick. That means, SBCs need to be deployed close the customers. So we have reintroduced the geographical planning complexity and the related capital expenditures.

This does not solve the problem fully. The access networks at either end could sufficiently impact the quality that having the SBC-based architecture may not be sufficient. By the time we address this new issue, I submit that we would have recreated PSTN equivalent from business complexity point of view. So we are better off sticking with the original concept end-to-end application, free of service providers.

Posted by aswath at 03:49 PM | Comments (0)

January 19, 2004

Carriers can not earn money on VoIP

So says Thomas Nolle in a recent column. He points out that “regulatory arbitrage” seem to the main motivator. This arbitrage condition will be eliminated once FCC makes its decision known, whatever the decision – either TDM service providers will artificially convert their traffic to IP to take the same advantage or VoIP carriers also have to pay the same termination charges. He also agrees that there is no money in completing on-net calls. So what is a company like AT&Tdo now? Interestingly, he thinks that AT&T can encourage its customers to migrate to multimedia IP calling and earn revenue. What is not clear is why multimedia calling will generate revenue? If VoIP is truly end-to-end, then why multimedia sessions are not? A session is multimedia because the end terminals are correlating multiple streams; the network at the IP layer is not privy to this fact. So the hunt for the business prospect in IP world continues.

Posted by aswath at 11:26 PM | Comments (0)

January 15, 2004

What is wrong with NATs?

Martin Geddes has an entry regarding NATs and its impact on end-to-end application. His final three sentences are jarring: “The market has spoken. Get used to it. Move on.” I would like to review some of the points he makes in his entry, a paper by Andrew Odlyzko that he refers and a paper by Blumenthal and Clark that Odlyzko refers. The question is has the market really spoken, if so what is its message and finally are the service providers hoodwinking the consumers into an architecture that is contrary to everybody’s expectations.

Martin notes that consumers like to use NAT because “it just works” for those users who use only web browsing and email. Others who want to avoid NAT just have to pay extra to get additional public IP addresses. This is analogous to business class customers paying extra unlike economy class passengers in an airline. As far as VoIP is concerned, SBC is a proven way to address the NAT traversal problem. Even though Skype does not say it openly, its solution is effectively based on SBC; except that the SBC function is implemented in fellow members’ (who happen to be in the public Internet) computers. So there are no insurmountable problems with customer owned NATs.

Now consider the case where the ISP is deploying NATs, meaning the subscribers “get a pre-NATted private IP address”. Andrew suggests that historically there are incentives for ISPs to move towards “price discriminate”. Blumenthal and Clark agree as well. The motivation seems to be the pot of gold called voice (is that why it used to be called “POTS”?). If this happens, then Skype’s solution will be in trouble because there are not enough members in the public internet to support the vast number of “NATted” members. In my reading this is the problem anticipated by Speak Freely.

Those ISPs who provide only private IP addresses to their subscribers are breaking the end-to-end model. Tough luck? No way. The service providers would like to keep the Internet unregulated because the technology fosters competition. But isn’t this an anticompetitive behavior? Aren’t they behaving like traditional telecom providers in the sense that they are artificially keeping their resources (IP addresses) scarce? This is the reason ISP based NATs are bad; not because it is ideologically incorrect. My appeal is that we have to ensure end-to-end nature is not hindered by the service provider, because this is the aspect that makes Internet competitive.

Posted by aswath at 01:23 AM | Comments (0)

January 13, 2004

Features offered by SIP Service Providers?!

As you probably know that there are many VoIP Service Providers. Whether they offer free service or charge a monthly fee they claim they support a long list of features. As somebody who wants to understand SIP architecture, we must evaluate how these services are offered. Today’s entry is an attempt at analyzing some of the features. (Even if you are a consumer, but otherwise uninterested, you are better off reading the full analysis.)

  • Call waiting: Since SIP is an “out-of-band” message oriented signaling protocol, it is simpler for the SIP Proxy to deliver the Invite messages without analyzing the state of the terminal. That means, the call waiting service is offered by the user and not the service provider. Many of them claim that call waiting is included in the monthly fee. This is not truth in advertising.
  • Transfer: This is another feature that is implemented in the user terminal; so it is not clear what is the role of the service provider.
  • Caller ID: The SIP Invite message carries the information of the caller; but this field is populated by the user. It is not clear whether the information appearing on the Caller ID display device is based on the original information or whether the Service Provider validates this information. If it is the latter, then the claim is justified; otherwise the claim needs to be diluted.
  • Caller ID Blocking: In PSTN, it is a simple matter to block the delivery of caller ID information. But in VoIP, it is not sufficient to block this information in the signaling domain. The RTP packets carry the source IP address which may need to be masked as well. That means there is a need for anonymizer function.
  • One can continue this analysis down the list. The service provider is needed for a small set of functions like advertising reach information, NAT traversal, 3-way/multiway conferencing. But in the rest of the cases there role is superfluous; worse the claims may be misleading.

    Posted by aswath at 05:46 PM | Comments (0)

    January 12, 2004

    Role of SBC in the VoIP Architecture

    Today, David Beckemeyer commented on an earlier post. Instead of replying him under that post, I am creating a new entry because the comment branches to a separate, fundamental topic.

    A clip of his comment is the following:

    One issue with application level gateways (ALGs) or SBCs, is that they fundamentally break a major benefit of SIP, that of separation of session establishment from session description. A SIP proxy isn’t supposed to know or care what the end-points are going to do, or how they're going to talk to each other. That's between the end-points. When a nifty new thing comes out, SIP proxies don't have to be updated to accommodate or support it. But a ALG that is handling media won't understand the nifty new thing unless it is upgraded or recoded.

    SBC is a nifty (if I am allowed to say myself) idea that comes at a price. To understand why it is nifty, a short description of SBC and its role in the VoIP architecture is needed. You may already know about them, but here is my version. SBC is strictly a layer 3 device, meaning it operates only at the IP layer. Specifically, it processes the header of a received IP datagram and maps the quad (Source, Destination IP addresses and port numbers) to a new quad. The mapping rule will be established by the SIP Proxy. With this simple device added to the network architecture, we can overcome the NAT traversal problem. Interestingly, we can offer additional services like Anonymous call, entry-point to a QoS-enabled network, and pivotal point for call altering services like transfer. SBC could be the point where Legal Enforcement Agency taps the media path. The basic function of SBC is already available in NATs (called Twice NAT?); but we need to add the control interface. This was one of the proposals to Midcom. I understand that Midcom is dormant; but given the activity in SBC market segment, it is safe to assume that SBC functions are required by the service providers.

    An observant reader will also note that the new architecture resembles circuit-switched network in a significant way: the standard circuit switch which is made up of a control element and a switching fabric is replicated by a Proxy and SBC. So it is very easy to port the service logic from the circuit switch to this new architecture. This is very attractive for traditional voice vendors and service providers. But it is easy to manage because one can add additional SBCs or can replace a failed SBC with a new SBC. No new calls will be affected even though existing calls will be lost.

    An SBC that follows this design (rather than the widely available kind that intercepts the signaling messages as well) does not encounter the difficulty that David is anticipating. But it does extract a price: since SBC is a funnel through which media traffic flows, SBC’s access link should be sized properly. The control interface between the Proxy and SBC must be protected.

    Posted by aswath at 05:44 PM | Comments (1)

    January 11, 2004

    VoIP Taxonomy

    David Isenberg, in his latest entry identified a news item from Yahoo News that categorizes VoIP into three groups: VoIP is used only as a transport technology bridging two PSTN segments; the second is VoIP begins and ends in the IP network; and finally the call traverses from IP network to PSTN. It looks like different regulatory regime will be imposed on each type and there might be a small problem with the current thinking.

    The second type may end up unregulated because it is no different than other data communication applications which are unregulated thus far. An inescapable decision. The first kind should be treated simply as a different transmission technology, just like when TASI or Frame Relay technology that were/are used in intercontinental trunks. It looks like FCC is also thinking along these lines. A smart move.

    It looks like FCC may allow free PSTN access to calls that originate in the Internet. Not a smart idea. How do we conclude whether a call originated in the Internet or not? I can buy one of many FXO-FXS converter boxes and make a call that originated in the PSTN to look as if it originated in the Internet. So the earlier smart move can be easily undermined.

    Bottom line: We can not determine the original access technology.

    Posted by aswath at 10:41 AM | Comments (0)

    January 10, 2004

    Vonage TI Pact: What Does it Mean?

    As you have probably read it by now, Vonage and TI announced that they are working together to provide VoIP equipment that is compatible with Vonage. I am not clear on the specific details, but I have my own interpretations and what it means for the industry.

    I am assuming that TI will have components that have SIP stack and FXS and packet GW functions. Additionally, I presume that a DSL/Cable modem built with these components can be easily configured to use Vonage; that is it can register with Vonage and use its services. With this premise, the following are my observations:

    • These modems are currently being distributed by DSL/Cable service providers, who will soon, if not already, have their own voice offering. Why would they allow Vonage’s SW incorporated in a product that they are subsidizing?

    • Currently many of the modems have router functions incorporated in them. I know of at least one DSL provider who has disabled these functions to avoid support issues. Indeed they specifically do not support home networking.

    • Finally, what will prevent a user to use this modem, but subscribe to another VoIP provider?

    In short it is not clear to me why and how Vonage benefits from this partnership. But the news that TI is integrating VoIP terminal and router functions (integration of modem portion is not critical) is good for the users. This will eliminate the NAT traversal problem and we come closer to the dream of true peer-to-peer communication.

    Please read alternate views on this news posted at GigaOm and VoIP Watch.

    Posted by aswath at 12:03 AM | Comments (0)

    January 09, 2004

    Legal Intercept and VoIP: Impact on the Architecture

    news.com has posted a story saying that Feds seek wiretap access to VoIP. It is not clear why they decided to publish the story today, even though Department of Justice, DEA and FBI submitted their comments to the FCC VoIP Forum on December 15th and has been available on FCC’s website since then.

    We still do not know what will be FCC’s position on this matter. Early indications seems to be that the classification of VoIP will be a function of whether there is interworking with PSTN is involved or not. Legal intercept requirement will necessitate change in the network architecture. There are two components to legal intercept requirement: the first is delivery of signaling information and the other is delivery of call (media) information. Since the service provider is necessarily in the signaling path, it is easy to meet the first requirement.

    To meet the second requirement, the service provider must be able to redirect a copy of the RTP packets to the monitoring agency. Since normally, the media stream does not pass through a network node, the end-points must be instructed to route RTP packets through a designated node. But there is a catch. The subject should not be able to discern that the monitoring is on. This means that all calls must be routed through a network node. This could be handled by what are called Session Border Controllers. Deploying SBCs is not a simple matter. The service provider needs to engineer the number of SBCs and must have sufficient network capacity to route the traffic.

    Posted by aswath at 10:54 AM | Comments (3)

    January 08, 2004

    ATA and Home Wiring - An Update

    If you are planning to connect a FXS device like ATAs and CellSockett, please read this article in Voxilla before procedding further. I received this link from SIPURA support team.

    Posted by aswath at 07:50 PM | Comments (0)

    Scaling back the goals at Skype?

    Skype co-founder Niklas Zennstrom recently gave an interview to BusinessWeek. For me each interview gives more information on their plans and I should say each is progressively scaled back version of the previous one, thereby eliminating conflicting statements in the previous interview. This interview is no exception.

    Other VoIP service providers are similar to regular telephone service providers, because they are “using the PSTN”. Huh? Another flaw in these companies’s business model is that they have huge cost associated with customer acquisition and billing. Skype on the other hand incurs zero cost for customer acquisition and of course zero operational cost. So how are they going to make money? They will make money by offering premium services like voice mail and conferencing. Mind you these are subscriptions (unlike Vonage didn’t you know), so the operational costs are low.

    One piece of information that I could infer is that both voice mail and conferencing will be software-based features. So there could only be one time licensing fee.

    Skype does not anticipate that it will replace the PSTN completely. It will coexist just as Fax and email complement each other. This is a refreshingly a realistic view not shared by some of the other VoIP providers.

    Posted by aswath at 12:43 PM | Comments (0)

    January 07, 2004

    Is this “wow”?

    BusinessWeek has a series of articles on VoIP. One of the articles discusses fancy features offered by VoIP. The article quotes an AT&T representative saying “VoIP gives customers a sense of wow”. So I eagerly read further hoping to see some interesting service, only to be disappointed.

    1. Sophisticated call forwarding: This is where an incoming call can be sent to several phones according to a predetermined algorithm. Of course we do not need VoIP to offer this service. It is being offered in PSTN under different names.
    2. Unified messaging: Again VoIP is not critical for this service. It is a function of the backend capability of the answering machine.
    3. Ability to dial from the contact list in the email: Here again, this has nothing to do with VoIP. It is a CTI function and could be done with TAPI capable software and modem.
    4. Correlating customer record and caller: This is another CTI function, not dependent on VoIP per se. I thought AT&T advertised this capability as part of ISDN. So what is new?
    5. Moves, adds and changes (MAC): If this is such a big deal then it is a shame (and shame on PBX vendors) because the technology wasn’t preventing from adding this capability. The cellular systems handle constant movements by dynamically identifying the base station associated with the user. In PBX application, we could replace the base station with the wall jack. Since the movement is less dynamic, a simple (software based) administrative procedure is enough.

    So I am still looking for features that really scream for IP. Of course the one I know is a killer application does not get much press – it is that two end-points can decide to communicate with each other without involving any intermediary. It is part of the famous highway analogy. To visit your house, we agree on my visit, sit in my car and drive to your house. I do not get concurrence from anybody else. Since IP allows me to do that, shouldn’t this be the premier feature? I guess, this wouldn’t be “wow” from the service providers’ point of view.

    Posted by aswath at 10:03 AM | Comments (0)

    January 06, 2004

    Yet another VoIP story in NYT

    Today’s obligatory entry should be NYT story entitled “A Debate on Web Phone Service” that appeared yesterday. Almost everybody who pays attention to VoIP related issues have commented on it. I have nothing new to add. Since this story has received such a wide circulation, I feel it is appropriate to repeat myself.

    The story points out that the cost differential between PSTN and VoIP can be attributed to differences in the regulatory regimes. So if the regulators make the same then this advantage goes away. With this advantage taken away, how to overcome the market inertia? It is not clear.

    It also says that startup costs are low for new entrants. It does not address whether this advantage is linear or tapers off as the network is widely and densely deployed. One of the advantages of VoIP over PSTN is that VoIP does not require geographical density. By this I mean, it costs the same to support 1000 users spread over 10 different sites or one single site. So a new entrant using VoIP will require lower capex. But I suspect that this cost advantage will diminish as the customer base grows.

    It is time to stop saying that packet switching is more efficient than circuit switching. These days straight PCM is the preferred codec. With 10 ms sampling rate and 60% activity level, a packet has 80 bytes of payload. With 40 bytes of overhead and 60% activity level, the required bandwidth is comparable to a standard call. Any way is this the era of cheap, infinite bandwidth. So why does this contribute to the cost structure. Don’t get me wrong; there are operational benefits like degraded, but usable service vs total shutdown during overload condition. But that is another matter.

    Before we address how to regulate VoIP and IP network access, we have to look at PSTN regulation structure. Even though PSTN is looked at as a voice network, I suggest that it is a converged network. (Huh?! It carries voice, fax and data (however clumsily).) But most of the regulations do not distinguish the applications; but focus only on the fact that a call used the “tele” network (I want to avoid phone in the name of the network). There are certain rules that are application specific. For example, there are rules regarding recording a voice call; but by definition there are record of fax calls. So PSTN could be viewed as made up of a transport network and an application component. Analogously, in IP domain, certain regulations could be placed on IP access and others on VoIP. For example, universal access does not apply to VoIP, an application; instead it is appropriate to levy it against access provider.

    Posted by aswath at 12:31 AM | Comments (0)

    January 05, 2004

    Will there be a “middle” in VoIP

    I am sure you have heard that unlike PSTN, in an IP network, everything takes place at the edges. Indeed it has become an old saw. This is true theoretically; but it is seldom true when VoIP is involved. Almost all the providers (whether they use MGCP, SIP or proprietary architectures) place themselves in the middle (of both the signaling and the voice flow) between the two end-points. It is natural that service providers need to process and act on the signaling messages. But why the media flow? MGCP architecture requires the media flow go through the Call Agent to handle supplementary services. Many SIP providers require media flow through their Border Session Controllers to address many issues, NAT traversal being the chief among them. Even Skype, which is billed a peer-to-peer system, uses such an intermediary node. These intermediary nodes may be at the edge from the Network layer point of view; but it is certainly in the “middle” from the application point of view.

    So what is wrong with having a “middle”? In the case of PSTN, it is not easy to change the service provider because 1) the point of connection to the “middle” is the network address of the user and 2) it is an elaborate and expensive process for alternate service providers to locate this “middle” node. On the other hand, in an IP network, the second point does not hold. Even though the network address and the “middle” node have no relationship, a service provider can use the application layer address as a vehicle to retain a subscriber. This is happening in the email service domain. Fortunately, there are ways for a user to overcome this and have user specific SIP address and have it mapped to service provider specific address, as long as calls are made in the native mode. For PSTN networking, we must ensure that users can carry their PSTN numbers from one service provider to another – that is Local Number Portability must be applicable to VoIP providers as well.

    Posted by aswath at 12:04 AM | Comments (0)

    January 04, 2004

    ATAs and Home Wiring

    About a month back a poster in LightReading posted that “VoIP will be a nice (sic, should be niche) product deployed as a last gasp by the RBOCs and it is not going to work.” The main line of argument is that products like CellSocket will allow users to use excess minutes from their wireless plan but utilize the phones already installed at home. I am hesitant to argue against predictions and so I am not very much interested in the voracity of this prediction. But in subsequent posts, it was mentioned that CellSocket suggests that one can “cut the cord” to the phone company and use any phone at home to place a call over the cell phone; whereas none of the VoIP ATA vendors suggest that their devices could be used in such a fashion. No explanation will satisfy the original poster because no published information could be supplied. I would like to suggest to at least the readers of this blog that even the existing ATAs could be used in the same manner as CellSocket., even though it is not clear to me why none of the vendors say as much in their website.

    Technically speaking, both CellSocket and ATA’s support FXS interface on the “line side”; that is they behave like a Class 5 switch, providing dial tone, collecting digits and generating ringing current. We just have to ensure that the telephone network’s Class 5 switch has been totally disconnected.

    The following rationalization may appeal to some: ATAs allow for phones to be directly connected. There is no practical limit on the length of the wire between ATA and the phone. The length could be extended by coupling two wires. This is exactly the arrangement when the ATA is connected to a wall jack and the phone is connected to another jack in the house. We have to make sure ATA is the only source of power on the line. That is why we have to totally disconnect the service from the telephone company.

    Posted by aswath at 12:18 AM | Comments (0)

    January 03, 2004

    What product is needed for the consumer market?

    Two kinds of VoIP products are marketed for consumers – stand alone terminals like Grandstream BudgeTone-100 and Analog Telephone Adapters like SIPURA SPA-2000. Based on the feedback given in public discussion groups, they are great first generation products. They are inexpensive and have great voice quality. Still they are deficient in a number of ways.

    The most critical of them is its inability to traverse the NAT provided by home routers; this requires special handling at a service provider through Session Border Controllers. The solution is very simple if these devices and the routers use UPnP.

    The second deficiency is that they seem to assume the consumers will be interested in abandoning access to PSTN networks. I am of the opinion that consumers will prefer to access all the networks that are available to them. In the recent FCC VoIP forum, James Crowe of Level3 remarked that one network can not be the strongest, but interconnected network is our strength. This means, that what is needed is an ATA that provides not only an Ethernet interface, but also Tip and Ring interface and a cell phone interface a la CellSocket, Incidentally, a multinetwork device like this will address the lifeline issues like E911 and power failure.

    Posted by aswath at 12:51 AM | Comments (0)

    January 02, 2004

    What is so special about Vonage?

    The title should be “Structural Separation – Is it Natural in an IP network”; but I decided to be fashionable and use Vonage, whether it is appropriate or not.

    It is conventional wisdom that structural separation encourages competition and that IP networking technology allows for structural separation. I have no quarrel The thesis of this entry is to suggest that one can approximate structural separation in PSTN and also one can block the benefits of structural network in an IP network as well.

    These days it is routine to point out that companies like Vonage offer innovative services and it is possible only because they are using IP networking technology. I would like to suggest that Vonage-like service can be offered even in PSTN. Imagine for a moment a new device analogous to ATA that Vonage uses. The only difference is the new ATA has a V.92 modem instead of an Ethernet interface. The user establishes a modem call to a designated number belonging to the new service provider and uses the data link to send and receive signals for a new voice call. Once the initial signaling is done, the modem session can be put on hold and the connection is used for the voice call. With this device, the new service provider can offer services just like Vonage. Since this is built on the standard POTS access, lifeline services will also be covered. I am not necessarily serious about this proposal. I am only suggesting that PSTN need not be vertically integrated as people tacitly assume. Structural separation is possible in PSTN as well. We collectively decided to get all the service from the local access provider.

    On the other hand, an infrastructure provider in an IP network can block Application Service Providers. For example, already there have been proposals to block P2P traffic. What will prevent an ISP to declare that VoIP traffic also belongs to this category? The reason I am pointing out this is to suggest that we should be vigilant that the structural separation is maintained. Yes, it is easier to maintain structural separation; but it is not forgone conclusion.

    Posted by aswath at 12:04 AM | Comments (0)

    January 01, 2004

    Fragmented VoIP Market

    The other day I came to know yet another free VoIP service provider (Free IP call) based on SIP. This started the following thought process. There are two kinds of SIP service providers – those who offer PSTN interconnection and those that do not. All of them offer free “on-net” calls. Usually there is a monthly fee for PSTN interconnection. Of course in most instances there are no interconnection agreements between SIP-based service providers. The last two points are of concern. Continuing maintenance of exclusivity by the service providers impedes Metcalfe’s Law. If the only source of revenue is PSTN interconnection, the service providers will be facing a strange anomaly – as their addressable market expands, the need for interconnection reduces, potentially reducing their revenue base. So it is critical that so called purple minutes are identified. We have to place an important constraint on purple minutes: they should require mediation of a service provider.

    Posted by aswath at 01:23 AM | Comments (2)

    Copyright © 2003-2009 Moca Educational Products.