Programmer Whatmaster?

I first became involved with the Internet naming business in 1997, during the initial switchback on the scenic route of my primary degree. This particular detour entailed taking a job under Dr. Niall O’ Reilly,“Niall” pronounced “Kneel”, as opposed to myself, who is “Niall” pronounced “Nile”. Hope that helps. then of UCD Computing Services as a “Programmer Hostmaster”. It was the first, but not the last time that I would have a job title that most of the rest of the world had never heard of.

Despite this nominal obscurity, the salary was a much-needed seventeen thousand Irish Pounds, and I had leave to attend various things important to resitting my exams while working on campus, so I was glad to have it – initially, at least.

As the title of the job implied, it came in two parts: the programmer portion was fine, but the hostmaster portion, despite the grandiose implications of the name, was not. “Hostmaster” has at least two distinct meanings, one of which was to do with Internet addresses, but the other one, the one under which I fell, was to do with Internet names.

What is this difference about, and why is it so important? Well, no-one except the Internet backroom folks really see Internet addresses, but more or less everyone sees Internet names. If you use the Internet at all – and if you don’t, I can only presume you’re reading this because you’ve discovered it lying around after the apocalypseIn which case: I’m sorry. – you type them every day:, and so on. The situation is that the names are really for people to use, and the addresses are really for machines to use. The addresses, which currently look a little bit like or slightly more bafflingly, fe80::22c9:d0ff:fe7b:18f3%en0, are how machines get data to and from each other. Indeed, as far as the machines are concerned, the addresses are the determining reality, and the names are just sugar on top for the weak-minded humans. The analogy that’s often used is the phone book (back when we actually wrote that stuff down on paper) – you’d have a name to look up, you’d do that in the book, and that would give you the phone number, which would allow you to make the call. In the world of phones the system for performing this translation is called the phone book; in the world of the Internet, the thing that translates between the names and the addresses is called the Domain Name System, shortened to the DNS.

As you can immediately see, a name like is a lot easier to use and remember and write down than e.g. 2a00:1450:400b:c01::93, so there is real value for any Internet user in having this easy-to-use name. Lots of value. In fact, so much value, that in the period from January 1995 to January 1996, the number of Irish domains – names like or – more than doubled as it became increasingly fashionable, then a competitive advantage, then absolutely necessary to have a website for your limited company, sole trader, statutory body, or unlimited liability partnership. The phone rang off the hook, and I, as the designated hostmaster, was the person who had to answer it. There were a lot of customers, and they all had very much their own set of opinions about the rules that pertained to registrations, and why those rules, in the main, did not apply to them. And rules there were aplenty.

For by the time I got there, a lot had already happened. The Domain Name System itself had been invented in 1983, by Paul Mockapetris of the University of Southern California and Internet godfather Jon Postel. (Before that, believe it or not, what was done was to copy around the single file that had every known Internet machine name and address from network to network.)Back when it was possible to draw the entire Internet on a whiteboard, that made sense; but very quickly it became impossible, and today that file would not only be gigantically huge, it would also require so many changes per millisecond no-one could physically type that fast.

The first entities to appear as Internet names were domains for commercial companies and for military institutions.See wikipedia. These unsurprisingly appeared “under” the top- level names .com and .mil. But there was a clear need to have top-level names that were not generic, i.e., were country specific, like .uk, .fr and so on. Being no stranger to controversy, Jon Postel had decided to avoid arguments about what exactly a country was, such that it would qualify for a top-level domain, by appealing to an ISO standard as the ultimate determinant of what country registrations were allowed.ISO 3166, if you must know; a subtle technique wisely employed. Accordingly, Ireland received it’s entry early in 1988.See wwtld for more.

At this point, we encounter a hint of the complexities which continue to haunt the .ie domain. At that time, the authority to register names was assigned to UCD (under the redoubtable Dennis Jennings) where in fact it still resides, as of the time of writing (late 2015/early 2016). Also, due to the poor connectivity in Ireland, it was not in fact possible to run the nameserversA term for machines that know about names, in this case contained within the .ie domain. in the country. Instead, Scott Bradner, a colleague of Jon Postel’s in Harvard University, ran the domain there. It was only in 1991 that it was moved from the U.S. to UCD proper. Dr. Jennings asserts that it was he who “called Jon Postel to ask that he give me .ie to run for the largely academic community”, and although I don’t have direct evidence to support this case – Dennis’s email archives prior to 2000 are unfortunately lost – it seems plausible. As I say above, it was certainly UCD who received this right, and as with much else in Internet governance, the authority to name flowed from nothing much else other than Scott’s ability to edit a file.

Securing registration sovereignty was a big step, but the early 90’s was still too early to see anything other than intermittent registrations. An email or two might have flooded in every other week or so. It didn’t take long for that to radically change, of course. By the time I came along, it was clearly necessary to do something more formal, and it fell to Niall O’ Reilly to do that something. Dr. O’ Reilly, in essence the General Manager of the registry at that time, was strongly influenced in his creation of the rules of the IE domain registry by the “fine job that Piet Beertema was doing with the .nl registry”.Personal communication with Niall O’ Reilly, Jan 2013. The Irish rules also influenced the Bulgarian registry when they were starting up.

Niall himself was deeply embedded in the European networking community, and had lived and worked in the Netherlands on secondment to the European Academic and Research Network, EARN, during which he met most of the prime movers for RIPE. It was natural for him to reach out to his colleagues in that community to help him. Ireland itself has generally punched above its weight in European institutions as a whole – perhaps leveraging our twin advantages of speaking English and yet not being English – and our small size and the comparatively small size of the technical community has meant that we have, sometimes, been able to move much faster than larger countries.This is most obvious in our head start in IPv6 deployments.

In the arena of naming, this meant that some reasonable subset of rules could be found for use without a round of expensive consultancy. Indeed, at the start of the Internet era in Ireland, that whole way of operating (expensive consultancy) was regarded as simply laughable at the time – why farm out to relatively ignorant consultants what you could engineer yourself in a few hours? I say this as an ex-consultant.

Although a not-unreasonable starting template, those rules were perceived by the public as being extremely complex and served mostly to stand between customers and what they wanted to do with the domain system, which was – in the main – profit from it. Although I had been known to have a gracious phone manner, after many months explaining to proprietors of shoes shops in North Kerry why they couldn’t have, what had been my previous breathy and welcoming tone became increasingly clipped, if not outright sharp. The number of customers who sensed intuitively that they were onto something potentially very big, if only they could be given the domain that they wanted, was very large. They were prepared to bluff, flatter, cajole or bribe in order to achieve their aims. With everything that was going on, I bought a stress ball to help me handle the constant flow of hostile interruptions, which – no joke – burst after a few weeks, and deposited a gummy sludge all over my keyboard. Another day, a medical equipment supplier turned up to the UCD computer center with a free sample of condoms for me, no doubt intended as slender compensation for my sticky accident, but that largesse did not avail him of a higher chance of

This understanding of the potential size of the opportunity amongst the public stood in stark contrast to the attitude of the ground troops of the registry itself. At that stage, few of us really understood then that we were sitting on a cash-generating business the like of which had little enough precedent in Ireland. It was an extremely resource-efficient way of making money, hand over fist. Gaze, for your edification, on the thing of wonder that is the table of domain-name growth from 1995 to 1996 below. Isn’t it beautiful? A business with monthly average growth of 7.3% – today, businesses that break 4% a year are really quite successful indeed.

Table 2: Historical Domain Registration Data from January 1995 to July 1996, from Niall O’ Reilly via author’s personal archives of IEDR-FORUM

Month End Registered Domains Monthly Growth Quarterly Growth Annual Growth
9501 347      
9502 370 7%    
9503 407 10%    
9504 433 6% 25%  
9505 462 7% 25%  
9506 498 8% 22%  
9507 535 7% 24%  
9508 571 7% 24%  
9509 595 4% 19%  
9510 620 4% 16%  
9511 676 9% 18%  
9512 739 9% 24%  
9601 766 4% 24% 121%
9602 850 11% 26% 130%
9603 839 -1% 14% 106%
9604 926 10% 21% 114%
9605 1028 11% 21% 123%
9606 1136 11% 35% 128%
9607 1224 8% 32% 129%

Look: the cost structure of the domain business is very simple. What you’re really selling is access to a website – without the simple name to type in, most folks won’t be able to use it. But you’re not selling the website itself – you’re selling the translation of the name into the address. So you need some servers, but not many of them, because the resources used by things that receive and respond to domain name questions – usually called “Domain Name Servers”, or DNS servers for short – are really small. For example, an average web page today is about 320 kilobytes,See this Google article for more. which translates into about 64 thousand words – a small novel. A DNS request is generally less than one kilobyte, say 200 words maximum. Therefore any capital or operational expenditure you make to run a DNS business as opposed to a website business is going to be much much smaller. Sure, you need some staff to register things, and run those self-same servers, more folks to take calls from the general public and so on, but it all scales pretty well if you put most of your registration activity through your website, which you’d be a damn fool not to do. So what you are charging for is, in essence, editing a small file to add, on average, about 12 characters for every customer. Other than that, it’s heating, lighting, and a lot of foreign travel.

Back to the question of the size of the opportunity. Part of the reason why this was not uniformly well-understood, and part of the reason why the registry was critically understaffed for so long, was because of the rules about what could be registered. If you scanned those rules and perhaps have registered a .ie domain in the past yourself, you will see how different they are to most things you see elsewhere – certainly in the well-known generic top-level domains like .com, .net or similar. It is true that those “namespaces” could be said to be free-for-alls, in which anyone with a credit card and a good ideaActually, scratch the good idea. could register whatever they wanted with nary a check on what right they actually had to that name.

A gold rush or land grab is the obvious metaphor for this now – indeed, a metaphor baked into the business model of many other namespaces since, since each new top-level domain creates the possibility of the set of all well-known companies and trademarks having to be registered again within that domain, effectively making a business for companies who do nothing but stare at their monitors anxiously waiting for someone to create a new domain.And lawyers of course. One thing the naming business has is a lot of lawyers. And money.} Although I cannot find any evidence of anyone saying so, it was as if at some point the criticism reached such a pitch that being different from much of the rest of the world became a badge of honour. It is probably incorrect to say it was clung to – identity is always a tricky question – but it was definitely viewed as a distinguishing feature back then, and still is today. David Curtin, CEO of the Irish domain registry, said in conversation with me that:

We did make a virtue of that difference. [Our advertising] talked about being secure and resilient and the being second safest in the world, and so on. For a period of time, that message resonated with SFA and IBEC and representative groups of small companies and then it waned, because they seemed to stop either caring about safety or that it just didn’t appear to be a big priority. (Consumers we know, certainly don’t care about safety.)

The emergence of the “managed registry model”, then, might well be a historical accident based on Dr. O’ Reilly’s friends of the time, but it became quite hard-coded from early on, and persisted for a long while. It should also be noted at this point that there are other claims to the motivation behind this emergence – Dennis Jennings in particular also asserted to me that all IEDR policies were rooted in decisions he made in the ’90s, and they were in general driven by the managed registry philosophy.

Whatever the truth is, it is clear that it was a determining component of the identity politics of the registry. Although the Irish approach had a lot of disadvantages from the point of view of the people who resold .ie domains,A quick note about how this works/worked. At the very early stages, there really wasn’t anyone else to get your domain from, so the IEDR served both as the regis_try_ and also as the only regis_trar_. Later on, when enough ISPs, hosters, etc had materialised, the IEDR was anxious to get out of the business of answering phones to customers directly and basically delegated the vast majority of its activity to accredited registrars – i.e. companies or organisations who would actually do all the talking to the end customer and just send the IEDR a list of domains to be added every so often. Well, that’s a simplification, but it’s ok for now. it must be admitted that it did in fact succeed in producing what was ostensibly desired,Although it’s not clear by who, exactly, or why. which was a very high-quality namespace where each domain had a very high probability of being traced to some reasonable holderAnother quick note on a foible of the domain-name business: because, amongst many other reasons, domain names could lapse and be re-registered by someone else, the registry did its best to say that the entity who was currently the registered contact for a domain was the holder of it, not the owner of it. of it. This is in some ways a net asset for the Irish nation, to the extent that we citizens (as notional co-”owners” of the .ie domain) are not contributing to the considerable problems associated with free-for-all registration.Including but not limited to trademark-related lawsuits, delayed product launches, concealed command-and-control networks for “botnets”, extremely large networks of e.g. home machines taken over by various means. It is also in some way a net loss to the Irish nation to the extent that business or other activities which could have operated within Ireland were lost to other jurisdictions.There are many examples of this: perhaps the most illustrative is the Information Society Commission in 1997, then chaired by Vivienne Jupp, who rang Dennis Jennings the day before they were launching to request a domain name. She was initially refused, which caused some consternation, until it became clear that the commission was established by statute to a degree sufficient to warrant a separate name. Well, Dennis made an exception on those grounds in any event.

It is also a loss in the sense that while the managed registry model as implemented in the early days provides some useful confidence that the registrant is legitimate, it blows the scaling model discussed above out of the water completely. The complexity of the rules and, consequently, the time and staffing effort required to enforce them caused the customers to complain, once again, about the costs. In fact, the costs of a name in .ie, particularly in the early days, were easily in the top ten in the world. One factor that pertained early on, for example, was the rule known as “one domain per holder”, or ODPH - this basically meant that one legal entity could only hold one domain name, thus forcing (for example) magazine publishers into creating additional companies for each magazine they held, or registering the unsatisfactory,, and so on.It was done away with in 1997. Further cost improvements happened later, such as the registry moving to a “presumed-ok” model, and having local access to the companies office database, but the cost question was probably not resolved fully until the registry had been 5-7 years out of UCD. There were quite a few other speed-bumps on the road to salvation too, as we shall see.

In any event, personally speaking, I resigned just when much of the firestorm of criticism was reaching what I then thought was a fever pitch in 1998, and went off to Ireland On-Line. I half thought that I would probably never see the inside of a domain registry again and don’t recall particularly regretting it. But fate was to perform one of its unexpected reversals that cause readers of mystery stories to wince at the implausibility and throw away the book in disgust. It turned out that what I thought was a fever pitch in this period was really just the most recent peak.

Programmer Whatmaster - December 1, 2015 - Niall Richard Murphy