Engnineering history is always interesting! My 50 cent response:

- IP Multicast was standardized in 1986, but it took until 1999 before it was commercially deployed for IP-TV over ADSL, and until 2005 before IP-TV services were widely deployed in my country for ADSL and fibre-to-the-home customers. The reason for the delay may have been that the telcos strongly believed in ATM instead of IP-TV and IP-telephony during the 90s, and that broadband internet access was not commonly available to homes until after the year of 2000.

- Ipv6 was standardized in 1994, and was implemented in Windows XP in 2002 and quite early in Cisco router equipment. But I did not see IPv6 addresses appearing in public web site logs until about 2 years ago.

- An early vision for third generation cellular systems was all-packet based communication, but 3G was based on double infrastructures - both circuit switching and packet switching nodes - just like 2.5G. It was not until 4G systems were deployed (LTE in 2009) when the circuit switching nodes were abandoned, and ip-telephony and an all-IP infrastructure were used in cellular networks.

- A lot of research has been devoted to advanced mechanisms for differentiated or guaranteed QoS in IP and mobile networks. In about 2005, QoS was implemented in fixed broadband access networks using a rather simple approach - by assigning different Ethernet VLAN-ID:s to different services, instead of using some IP level approach. I donät know to which extent differentiated QoS is implemented and utilized in today's mobile networks.

You asked about present ideas that may wait. Mobile TV services are still not very successful, for example DVB-H, 3G MBMS (multimedia broadcast multicast service) and 4G eMBMS (enhanced MBMS). But it is hard to now if multicasting and broadcasting over wireless ever will attract service providers and customers, or if the new generation will prefer video-on-demand over unicasting instead (youtube and TV play services).

Thank you Magnus for your response which is "a 50 cent response" as you say but multiplied by a million. All issues you raised are of interest given with a good explanation for the reason of the delay. I think that, out from your response we can say that in many cases the less-interest of businesses, as in the case of All IP 3G and IP multicast can in many cases cause the delay. In some other cases, some technologies will have to attract customers' interest like Multimedia broadcast services. For the IPv6, I think that it was put ahead of real demand and the need for it is coming know with the advent of sensor networks and the Internet-of-Things Thanks. @AlDmour

Thank you Ismat for the encouragment!

You mentioned UWB. I wonder why UWB and wireless-USB still are not commercially successful. (Actually, one of the suggested radio-interfaces for UWB is based on OFDM, i.e. a very large amount of carrier-modulated signals, rather than spark transmitter impulses, but still it is ultra-wideband just like the Hertz 19th century experiments.)

The delay of IPv6 was also caused by the introduction of NAT - an ugly work-around.

A related question: Can someone give examples of unsuccessful technologies that the research community did believe in, and explain why they failed?

I believe some technologies were developed way too early, and were made too advanced. They were technology driven rather than market driven. For example, DAB (the digital-audio-broadcasting system Eureka, or "dead and buried"...) is a rather complicated standard as compared to for example DVB-T, and DAB receivers were really expensive for many years and because of that, it never became popular except perhaps in the U.K.

Also, some technologies were designed for hardware implementation rather than with software in mind. Hardware is efficient but not very flexible.Several competitors to IP were hardware oriented and rather complicated and inflexible, and lost the battle in the end. For exampel X.25, Frame Relay and ATM.

Often the first technology will win the battle. IP (1974) was a couple of years ahead of X.25 (1976).

I think Low-density parity-check code (LDPC) were invented in the sixties, but considered too complex to use at the time. They had to be re-discovered and are now widely used.

Thank you Mattias for the good example of something forgotten for some good time because of its apparent complexity and then re-discovered to be beneficial regardless of its complexity.

Thank you as well Magnus for the expanded comment and the suggestions. For UWB of the spark experiment, I think the closest to it in modern communications is the Wide-Band CDMA (Spread Spectrum) kind of communications in which the signal is spread over a large bandwidth of around 5MHz. Of course, it seems that UWB based on concatenating narrow spectrum chunks (e.g. based on OFDM) is a more efficient means of communication because of its flexibility (flexibility in sharing the spectrum, assigning BW based on throughput needs and the ability to control ISI). Your question "Can someone give examples of unsuccessful technologies that the research community did believe in, and explain why they failed? ". is a nice alternative. Thanks. @AlDmour.

Shucks, I was going to use the LDPC example.

As to ATM vs IP, I think the fundamental problem with ATM is that it becomes impractical in large deployments. ATM requires state information to be retained at each node, for every "virtual circuit" transitting that node. The beauty of IP is that it is a connectionless, packet-switched best effort protocol, where each packet is routed indepenedently of the others in any given session. The intervening nodes, in the most typical case, don't need to know anything about that session.

IP multicast actually has this disadvantage, compared with IP unicast. With IP multicast, the nodes do need to retain information on whether or not there are any remaining members of each multicast group they route packets for. Another disadvantage of multicast is purely economic. The ISPs have a harder time knowing who to charge for bandwidth usage.

UWB is an excellent example of something that did NOT work out as expected. I think it failed for a couple of reasons. First, because at least one of the two types attempted to force-fit OFDM into a UWB scheme, which seems really counterintuitive. OFDM is not very good for low peak to average power ratio (PAPR), and yet UWB depends on low peak power. But maybe a bigger disadvantage has to be that high frequencies propagate less effectively than low frequencies, at equal power. Therefore, a scheme which depends on ultrawide band will suffer as range increases, simply because those high frequency components will attenuate very quickly. So now instead, everyone's looking at very high carrier frequency communications, in the many 10s of GHz, where more traditional channel bandwidths are used (channel width compared with carrier frequency), but with very high capacity thanks to that high freq carrier.

Thank you Albert for the comment. In fact some ideas which seem to be going away some times only hibernate and may flourish back at the right moment. For example circuit-switching technology is not generally preferred for networking and packet switching is the norm. However, it is realized that circuit switching will not go away fully, circuit switching will be needed whenever it is more efficient to do so, e.g. for large bandwidth connections such as those connecting data centers. Thanks. @AlDmour.

Hi all,

just to insert my favourite example for a communication result that has not yet been adopted but is currently in the pipelline for wide adoption:

Linear Network Coding and Random Network Coding for Graphical Multicast Networks:

- Based on Max-Flow-Min-Cut from 1956

- First proven that the cuset bound holds for multiple destinations with linear network coding in 2000 (Ahlsvede, Cai, Li, Yeung) and refined in the last decade. However, it has not found wide adoption. Fortunately, random network coding it is a candidate for 5G Device-to-Device (D2D) communication...

Best regards.

Thank you Martin, nice example.