Letters To The Editor

Ed.: We received several letters in response to the article “A Retrospective: Twenty-Five Years Ago,” by Geoff Huston, published in the previous issue of this journal. Here is some of the feedback:

Hi Geoff,

Just wanted to show my appreciation for your nice article. As an ex-DEC who moved to WorldCom after my MSc in Computer Engineering & Telecoms with a Master’s project on IP signaling over ATM, I can certainly relate to a large part (not all ;–) of what you wrote.

I normally don’t read such long articles, but had to make an exception as I kept interested until the end!

Thank you!

Pedro Paiva, Etoy, Switzerland



Greetings Geoff,
I just wanted to let you know that I really enjoyed your recent article, “A Retrospective: Twenty-Five Years Ago,” published in The Internet Protocol Journal. I lived through most of the history that you talked about as I came up through the telecom industry and then finished off my career at Cisco.

It certainly is interesting to reflect back on all the past controversy around network infrastructure design and how competing ideas and philosophies played out. (Talk about losers, remember Switched Multi-megabit Data Service (SMDS) driven by the Regional Bell Operating Companies (RBOCs)? While at Nortel, I remember once in a design review meeting that one of our BNR geeks put up a slide (overhead foil back then) that showed various network evolution scenarios. The last one was an “oh-by-the-way, there’s this theory that the Internet could take over the world” (of network infra- structure). All the room snickered. Who’s laughing now?

There was as much energy, maybe more, put into defending architectures based on market control as there was on technological elegance. Still, it is a fascinating and dynamic industry full of extremely smart people with clever ideas, and I enjoyed every minute of it.

I started at “the phone company” in the late 1960s and it has been quite a journey from relay-driven switches controlling tip and ring loops to the current Multiprotocol Label Switching (MPLS) back- bone networks, terabit switching, and hitching rides on photons.

Thanks for your insight and for your well-written article. Best regards,

Marc Williams


The author responds:

Hi Marc,

Thanks for your note and your recollections from some 25 years ago.

I recall SMDS as well. If I recall correctly, this was an invention coming out of a university in Western Australia. Elsewhere in the world it was marketed as a 34-Mbps product. In Australia it was marketed in 2-Mbps and 10-Mbps forms (evidently the telco thought that we primitive Aussies were not “ready” for any higher speed!). I was a customer of their 10-Mbps product, and experienced some disappointment when it became evident that 10 Mbps was a theoretical peak that was simply unachievable because the inline PCs that were used for packet accounting slowed the throughput of any SMDS link down to just 3 Mbps! So in Australia SMDS was largely killed by the telco and it was never really used for high-speed digital trunk services.

I experienced a similar reaction to the Internet in the late 1980s as you have observed, when, in response to suggesting that the universities were about to build a national IP network, many of the telco managers did the polite snicker performance and then suggested that we should “get with the times,” sign up as customers of their national ATM network, and leave the engineering to them. I’m glad the universities saw through it and supported me in persisting along the path to a national IP network. It was a strange moment some 6 years later when the same telco came knocking on our door to make an offer to buy the network from the universities because their own efforts to construct an IP product were simply getting nowhere at the time.

It has indeed been quite a journey, and I too have enjoyed every bit of it!

Kind regards,

Geoff, Chief Scientist, APNIC



Hello Geoff!

I haven’t chuckled that much in years; what great memories. A few of my strong memories:

  • Lack of documentation for new functions in software required an off-net test network and a Sniffer. The amount of hours spent figuring exactly what the function was doing or wasn’t doing could fill an ocean. Absolutely my favorite activity and still is.
  • I inherited a stat-mux system that was transporting ASCII terminals back to a centralized DEC terminal server arrangement. Hated it with a passion. One day, after a couple of beers, a light bulb came on that Ethernet is a stat-mux, so I bought a couple of Cisco AGS units, remotely installed a terminal server and an AGS, hauled it back to the other AGS in the central location and danced a jig, and then I started ripping out the old WAN stat-mux the following week.
  • Anything relying on a token for timing is pure evil. You never know when you’ve engineered a TTL exhaust until it happens, and that can be based on Distance + Nodes or pure application coincidence. Ring resets are the devil’s work. Token-based systems are not stat-muxs, but Ethernets are; that’s why Ethernet survived and is the “last man standing.”
  • I totally agree with your comments surrounding the “cloud.” I can remember that the distributed-versus-centralized fad has occurred at least four times over the past 25 years …
  • Z80: I built my first PC with a Z80; thank goodness for the peek- and-poke function!
  • OEM would claim anything was portable as long as it had a carrying handle attached, even if it took two people to carry it.
  • I fell in love with TCP/IP very early for the simple reason that it has the best of both worlds: a tightly coupled connection and connectionless protocol. It is much faster to troubleshoot or modify because IP requires a different expertise than TCP, and when you run across individuals who can work across the layers, hire them!

So, a lot of fond memories. I started out as a telemetry engineer on the Apollo project and I thought that was challenging and fulfilling. But, it doesn’t hold a candle to the 1984–1995 period.

Oh, one other thing; I take umbrage to “…the annoying persistence of FORTRAN.” That’s the first language I learned back in the late ’60s and I still have an active compiler on an old laptop that I still program on … LOL!!

Keep attacking the certificate situation! The current situation is a disgrace, and I fully support the concept presented by Barnes: let’s hurry it up!


Paul Dover


The author responds:

Hi Paul,

Thanks for those recollections. I too spent a massive amount of time staring at a protocol analyzer, trying to make an IBM PC look enough like a Uniscope to allow file transfer between the PC and the Univac mainframe—no doubt it was a character-forming experience, but all I can say now is thank goodness for tcpdump and wireshark!

Thanks for your note—I truly appreciate the feedback!

Warm regards,

Geoff, Chief Scientist, APNIC



Dear Ole,

Congratulations on your 25-year anniversary!

You can tell how well people enjoy their professions by how great their products are, and yours is in the “excellent” category.


Paul Dover




Congratulations on your reaching a major milestone: 25 years of technology publishing! We are glad that you are continuing this service through The Internet Protocol Journal and look forward to many more years in this field.


T. Sridhar

Leave a Reply