Fateful Choices, Unforseen Outcomes

We often do not see clearly enough that when the global communications industry chose TCP/IP as its next-generation network architecture, it also chose a business model based on layers, permissionless app development and ownership, an open ecosystem rather than a closed pattern. 

That was perhaps not intentional. The bigger drivers were the relative cost and ease of operating a network using IP, rather than the rival asynchronous transfer mode network, its main rival as a solution. 

In doing so, connectivity providers decided all networks would be computing networks. But where is value created on a computing network? By users, at the edge, running applications deemed to provide utility, connecting buyers and sellers, trading partners, friends, relatives, colleagues. 

The network itself is a functional “dumb pipe,” supplying transport. That is essential, to be sure. But value lies in “ enabling connections to be made.” Beyond that, the network is not essential for computing functions, application development or deployment. All that happens “at the edge.”

That’s just another way of saying people, companies, institutions or groups can create value, use cases and apps that use communication networks without the permission of the network owners. That is the complete opposite of the case before IP networks were the norm. 

It was inevitable, if largely unforeseen, that most applications and network-delivered products could be, and would be, created and owned by third parties not under the ownership and control of connectivity providers. 

These days, nearly all communications over public and private networks are connectionless, a far cry from the situation four decades ago, when most sessions were connection oriented. That is another way of saying the way we communicate now uses packet switching.

source: TechTarget

Connection-oriented protocols were characteristic of the telephone voice network. Connectionless is the way modern data networking and internet protocol work. 

In that regard, and with passage of more than two decades, it might be worth noting that all global public communication networks operate on the same principles as computer networks: connectionless. 

Another way of saying this is to note that communications are essentially edge to edge, and not deterministic in the core transport network. We might also say such networks are essentially dumb, not “smart.”    

More than two decades ago, some debates were held on the merits of “smart” networks versus  stupid networks. A smart or “intelligent” network was touted by those who believed it best supported a range of applications where predictability and therefore quality assurance was important. 

The “dumb” network was preferred by data professionals, since it was the way computer communications occurred. Back then, the general framing was whether either IP or asynchronous transfer mode was the better choice for wide area and public network communications. 

Local area networks might have used routers or switches to connect LAN segments. So, as applied to the WAN, telcos proposed adding an ATM transport function, partly for quality assurance, partly for capacity. At the time, ATM could support faster speeds than IP could (though obviously that changed). 

 Data networking professionals asked “why?” Simplicity is more valuable than determinism, they argued.  But all those debates were over technology choices. 

source: Shanker presentation 

The debate seemingly over technology was much more than that, though perhaps little understood at the time. We might argue, persuasively, that the global IP networks are precisely “stupid” or “dumb” networks. In the IP framework, the transport network routes packets. 

In that sense, smart network elements are essential, and used. But all applications are created in an independent manner, even when directly owned and supplied by a connectivity provider. That is simply the way software works these days. 

Connectivity providers can build features that are more deterministic, such as layering on multi-protocol label switching to add more predictable transport quality. But so can edge devices owned by third parties or enterprises directly. 

Virtual private networks can be created to enhance security. But those features can be created either by transport providers or edge hardware and software owned directly by  enterprises. Service businesses can create such networks and make them available to consumers as well. 

Those entities can include--but are not limited to--transport service providers. Even now, most VPNs are created and sold by third parties, not transport service providers. And much of the activity by service providers are entities separate from the major transport providers. 

So layers, open, permissionless, connectionless, disaggregated, at the edge, over the top and dumb pipe are some of the common realities of the modern computing and connectivity functions and industries. The transport function ideally is transparent to all the other layers and functions. 

But that also has other implications. The value of “communications,” while remaining essential, arguably gets devalued, as sources of value move to the edge, and away from transport layers of the full solutions stack. 

More than two decades after a ferocious attack on the notion that the best network is a “dumb pipe,” isn’t that precisely what we now have, for the most part? 

Granted, one can still argue that service provider apps such as voice and text messaging or video entertainment or home security are applications that still require some amount of “intelligent” control in the “core” of the network.

But most of the bandwidth now carried on any network consists of internet or other Internet Protocol traffic, by definition edge-to-edge routed load. 

Even two decades ago, we might have all learned faster if the debate had been about connectionless versus connection-oriented networking. `We might have gotten a better grip on what a “layered” architecture  would mean for business models. 

We might have seen the “over the top” business model coming. We might have foreseen what “permissionless” ability to create products and reach users and customers directly would do for, and to, business models across the ecosystem. 

We might have gotten glimpses of new models for value creation and hence monetization. But what seems quite obvious in retrospect was anything but clear at the beginning.