Netheads vs. bellheads redux: the strange victory of SIP over the telephone network

How much do you know about how the PSTN (that is, the public telephone network) works? At some point long long ago, it was signalled using in-band frequencies, which facilitated the infamous phreaking using so-called “blue boxes”. This led to a switch to out of band signalling, ultimately culminating in something known as Signalling System 7 (SS7). SS7 traditionally was carried over things like a T1/E1 or similar. The standards for SS7 are defined by the ITU-T and have basically no “cultural” relation to the Internet or IP. Ultimately, it defines a circuit switched communications network based around separated signalling and time-division-multiplexed (TDM) voice channels.

And then, of course, the internet happened. This led to an interesting conflict between two fundamentally different cultures of networking and associated visions. This has been popularly documented as the “netheads vs. bellheads” phenomenon. The difference between these worldviews is fundamental. The bellhead view saw communications in terms of circuit switching, and viewed applications as being integrated into a smart network and provided by the telco.

It's actually interesting to note just how oblivious the bellhead world was to its ultimate fate. Once upon a time, I flipped through some old books about ISDN out of idle curiosity. One of the striking things was the repeated use of the term “datacom”, a term which now seems basically extinct. The reason I found this striking is because it implies that, to those working in the telco field at the time, “datacom” was not the primary purpose of a network but a kind of strange aside, a peculiar niche addition. From the modern perspective this worldview seems almost comical; ironically, the term datacom is no longer relevant because everything is “datacom”.

Of course, to the extent that the bellhead culture did have awareness of what was on the horizon, the connectionless communications paradigm enabled by packet switching was viewed as a threat to their revenues, as was the idea of a dumb network which eliminates the possibilities for telcos to provide value-added services.

There was a period of time where telcos were trying to fight this by providing new value-added services; it's weird to think that just as recently as the early 2000s, 3G cellular service was new, and boasted the ability to pay your carrier an exaggerated per-minute rate for the ability to make video calls to people, with a level of quality which now seems laughable. Ironically, you are now less likely to have to pay for a video call (which is likely to be made over a platform like Zoom or similar which certainly doesn't have anything as absurd as per-minute billing) than a regular phone call, which might still go over the PSTN. Of course, much more discussed is the attempt by telcos to fight their commoditisation by undermining net neutrality in various ways. There was a time when people were seriously concerned about the destruction of net neutrality and what it would bring. People's fears of a cable TV-esque model of internet access were communicated in this widely circulated memetic image. These fears now feel as though they are in the rear view mirror. The telcos lost. Once upon a time, some cellular carriers actually tried to contractually prohibit the use of third party messaging services over IP, lest you undermine their SMS revenues. These times now seem a distant memory.

So it occurs to me that it's significant, yet under-remarked, just how utterly the netheads have won. This directly parallels my previous article about “The cultural defeat of Microsoft”, which is basically the same phenomenon but in relation to the utter triumph of UNIX-like operating systems over Windows.

By “cultural defeat”, I mean a whole way of conceptualising a class of technology, like computer networking or voice calls, or how to arrange the design of an operating system. UNIX and Windows NT (being a fusion of DOS and VMS) constituted different cultural domains, whereby the workings of the other were mutually alien to those used to working in either. You would have developers who lived their lives in the Windows ecosystem to whom UNIX might seem alien; the same was true in the reverse. The important thing to note is the existence of a viable ecosystem of developers living the Microsoft way of doing things.

To be culturally defeated as an ecosystem is to have this cease to be the case. The cultural defeat of Microsoft can be no better demonstrated than by Microsoft's creation of WSL in an attempt to appeal to developers. Microsoft no longer has the mental dominance among developers, or even a sufficiently significant minority of them, to be able to dictate an independently viable culture; instead, they must appeal to the crowd to whom UNIX is the normal way of doing things. In defeat, they must even accept the UNIX OS onto their own platform. Microsoft's use of Linux internally is also inevitably extensive, thus the degree to which they even dogfood, and thus practice, the culture of their own OS will continue to wane.

This same process has happened with the bellheads. The PSTN, based around SS7 and ISDN, is a Bellhead design through and through. It also no longer exists. We continue to entertain the strange pretence that the PSTN is still a circuit-switched network, but it isn't. The interface presented to the end user may still seem unchanged, but the entire thing has been hollowed out and replaced internally with VoIP. Even where SS7 is still used, it's often tunnelled over IP rather than actually running over the traditional TDM links. Indeed, the only thing that remains “circuit switched” about the PSTN today is the per-minute billing model — based on telcos mutually pretending to one another that they're still operating a circuit switched network that justifies this kind of billing.

What I find interesting though is the sheer extent of the victory of the netheads. You see, there are many different VoIP signalling protocols. The most commonly used VoIP protocols are SIP and H.323. It's not inaccurate to say that SIP is the nethead VoIP protocol, and H.323 is the bellhead VoIP protocol. (In fact, it's an under-remarked fact that H.323 is almost literally just the ISDN signalling protocol running over IP.)

These distinctions aren't exaggerated. SIP is a textual protocol resemblant of HTTP, email headers, and the like. Moreover it's designed to support federated calls between internet domain names, just as email is federated by domain names. Theoretically SIP could have been deployed on the public internet in much the same way email is today, with email addresses reused as SIP identifiers, meaning that you could call an email address directly and without contracting with or paying any intermediary. SIP is a product of the Internet culture of protocol design through and through.

There are probably many forces that pushed bellhead telcos towards VoIP based architectures internally. It's not hard to imagine the most significant may have been cost; as demand for internet bandwidth increases and PSTN usage dwindles by comparison, sourcing the relatively bespoke equipment to handle TDM links and SS7 obviously becomes more expensive, as does finding people raised in the workings, operation, and more generally the culture of these technologies.

Yet when telcos adopted VoIP, they had a ready-made solution which preserved the essentially Bellhead worldview of telecoms: H.323. As far as I can tell, usage of H.323 was actually more prolific earlier on, probably because H.323 was first to market, coming out as a standard several years before SIP. Older versions of Windows had a video conferencing program called NetMeeting which was actually H.323-based.

Yet if we fast forward to today, we find that SIP is entirely victorious. Carriers interconnect using SIP. H.323 seems practically dead. This means that telcos now operate the still-pretending-to-be-circuit-switched PSTN using an utterly nethead technology running over nethead network technologies running over nethead physical interconnects. To my mind this is deeply interesting. While telcos may have been driven to use commodity IP hardware to lower costs, the early dominance of H.323 means no comparable force pushed them towards SIP. A priori, I would expect telcos to be more fond of H.323 due to the closer correspondence of its semantics to the PSTN's existing workings (as, again, it is almost literally just the ISDN signalling protocol run over IP). Interworking between H.323 and the PSTN almost by definition necessarily involves fewer “impedance discontinuities” than interworking between the PSTN and SIP. SIP's nethead, URI-centric design means that working the PSTN over SIP is slightly unnatural and requires that telephone numbers be converted to URIs such as sip:+18005551234@example.com, but doubtless there are many more slight discontinuities.

In fact, the victory of SIP as the technology used to sustain the enduring illusion of the PSTN is now so total that as far as I can tell, it's essentially the law in the US. The FCC has mandated the adoption of STIR/SHAKEN to try and stop robocalling. STIR/SHAKEN authenticates inbound caller ID and is essentially to SIP what DKIM or SPF is to email. Moreover, STIR is an IETF standard pertaining solely to SIP.

As far as I can tell, you can still run a TDM network in the US. However, there has as far as I can tell been no effort to extend SS7 to be able to transport this authentication information. Thus any call which passes through a TDM network has no authentication information and is considered suspicious by default, similar to an email without a DKIM signature. Thus while TDM networks can continue to operate, increasingly they may find the calls they originate discriminated against in comparison to calls routed via SIP. The writing is clearly on the wall.

There appear to be some more recent efforts to develop an “out-of-band” variant of STIR/SHAKEN which is intended to try and accommodate the remaining TDM networks. This appears to be based around the idea of allowing the authentication information to be registered with some kind of server before the call enters a TDM segment of the path between caller and callee, so that once things come out as SIP again on the other side, the authentication information can be retrieved from that database. What is striking here is the implicit lack of any attempt to actually extend SS7 itself; there is an implicit assumption here nobody is really interested in any attempt to actually further develop SS7. The technology is a dead end — and aside from the now dwindling H.323, only nethead technologies are available to replace it.

Further examples of the victory of the netheads are found in the “all-IP” architecture of 4G and successor cellular standards, with voice functionality being provided by the SIP-based VoLTE. This is an even more interesting case where telcos are seemingly unable to resist the advantages of using commodity networking technologies and a unified architecture. The desire to charge for things like video calling is gone; in the era of Whatsapp, carriers seem to have come to a state of relative acceptance about being a dumb pipe, and in some ways now embrace it.

To sum up, we started with the world of bellheads, with a dominant infrastructure and culture, and even bellhead VoIP technologies like H.323 which had a substantial head start over SIP. Yet it is now nethead, IETF-developed standards which form the basis of all of these telcos. The only aspects of the bellhead culture and technology stack that seem to remain are the user interface aspects; the generated illusion that what you are using is some kind of circuit switched network.

Moreover, organisations like the ITU-T increasingly seem to act as patsy standards organisations to the IETF in the same way the W3C does to the WHATWG. As I mentioned in my article on XHTML2, I characterise the creation of WHATWG as essentially being a kind of coup in which the effective authority for setting web standards was transferred to WHATWG. What is interesting is that the W3C, facing the prospect of its own irrelevance, has essentially chosen to save face by still publishing web standards, but they always happen to be whatever WHATWG published. The illusory nature of the supposed authority of the W3C is illustrated by the rare cases where the W3C actually tries to make their version of a standard differ from that of the WHATWG's, followed by them backing down and getting into line.

In the same way organisations like the ITU-T essentially have to maintain their own relevance by accepting the reality that the world of telecoms is now based on IETF protocols, which as far as I can tell it does by publishing profiles and architectures built on IP networks. Based on the examples both of the W3C and the ITU-T, it seems like when a standards organisation experiences a sudden loss of authority, the response is often to save face; this at least allows the organisation to continue existing and to try to justify its existence.

The ITU-T approach to protocol design of course remains available for viewing in their older standards: things like the failure of ISO OSI or the use of ASN.1 (but for some remaining use of ITU-T technologies, like X.509, which is basically solely due to those technologies being adopted by the Internet community). There are actually interesting things about how the ITU-T sought to approach protocol design, such as the use of domain-specific languages for modelling protocol behaviour (SDL, MSC, TTCN-3; LOTOS, a temporal logic language; Estelle; and of course ASN.1).

What do I make of all this? As I've previously written about before, I'm not fond of the telephone network. Having said that, I will say I do like to see diversity of cultures in technology, and seeing a completely different way of doing and thinking networking collapse into a monoculture makes me a little sad. I'm very happy the netheads won, of course. You could perhaps say that reading about TDM networks, SS7, ISDN, and ISO OSI is a bit like discovering the ruins of an ancient civilisation. It may not have been a civilisation you would have wanted to live in, but you can't help but feel sorry all the same.

Moreover, while the netheads did win in terms of the adoption of network technology, the victory was not total. The original vision of SIP, of calling between internet domains email-style, has never been realised. The modern PSTN runs on SIP, yet it is almost without exception operated privately, exclusively between parties who have formed a contractual relationship as the basis for interconnection. This interconnection is often done via private IP networks rather than over the public internet. Voice calling between domains without prior contractual relationship, as is the case for email, is nowhere in sight. Paradoxically, telcos have replaced all bellhead technologies with nethead technologies, and then used them to continue to sustain the illusion of the same service as was provided using bellhead technologies (right down to increasingly ludicrous and unjustifiable interconnect fees and per-minute billing). The most prolific use of the nethead SIP technology, strangely enough, is now to prop up the simulation of a bellhead network.