It has been nearly two decades since there was any serious debate in telecom circles about the choice of a next generation network, when the respective roles of internet protocol and asynchronous transfer mode were evaluated.
The debate was resolved in favor of IP. Few decisions ever made in telecom have had greater implications. The reasons were almost stunningly simple: ATM proved to be too costly; less flexible; less compatible with the growing need to support IP-based end user devices, apps and services.
The biggest change is that where support for voice had been key, the new requirement was support for computer communications. In essence, the network was to become one giant computing network.
Traditionally, the debate had been about how to ensure quality of service. That was the ATM advantage. QoS for computing communications was a completely separate matter. But it seemed obvious enough that future networks would have to be optimized for computing, not voice, in any case. That clearly was the right call.
Quality of service therefore emerges as a key issue for some applications, especially those which are highly susceptible to packet arrival delays. In choosing IP as the next generation platform, service providers embraced the data future, at the cost of creating some potential issues for latency-sensitive apps.