At the risk of stating the obvious, I’ll point out that the analogy falls down at a critical point. Whereas two independent navigation systems can coexist without conflict, two independent domain name systems will naturally contradict each other unless carefully coordinated.

OK, two different types of positioning systems. Fine.

Different encodings for TV signals (PAL, SECAM, etc).  Fine again.

They can work just fine and dont interfere with each other.

Now could you please stop trotting out those bad analogies to try and prove your point that two sets of rootservers can coexist?

To Bret:

We seem to agree that the real concern is consistency of data not singularity of source of that data.

My own feeling is that there is a natural coercive force that will drive consistency as well as a legal force.  The natural force, or at least an economic force, is that people don’t like to be surprised and they will shun a DNS system that gives answers they don’t expect.  That will, in turn, drive providers, through their own perceived self interest, to be consistent, at least in the core TLDs.  The legal force is that of intellectual property law - someone who believes they have established an enforceable trade or service mark will have at their disposal the full panoply of legal tools to redress and potentially supress the offerings of others who use the same TLD string.

There is, of course, the issue of size.  The EU is bigger than the US, as is China.  So if they, individually or together, form a root system, then who is to say, as between the one run by the US and these others, which is “authoritative”?

The point that I am making, however, is not about consistency but rather about the fact that despite there being a workable global syste (GPS) there has been perceived value in establishing the separate Galileo system.  The same forces that drive nations to create a parallal and redundant global positioning system also exist to drive nations to create a parallal and redundant DNS root.

As for the other person’s comment:

It seems to me that the shoe is on the other foot, that those who want to violate the end-to-end principle and eliminate user choice ought to bear the burden of demonstrating that their desire is of such supreme importance and of such an overiding and compelling nature that it warrants the supression of individual choice on the internet and the wholesale abandonment of the end-to-end principle.

And that is a burden that those who oppose competing, consistent systems of DNS roots have failed to undertake, much less demonstrate.

Moreover, we already have several operational systems of roots.  Some we have had them for years.  Others are coming online to meet the demand for internationalized names.  Yet the net has not collapsed.  And if the net actually were actually in such a danger of collapse simply by people setting up their own DNS root then one must wonder why such a feeble technology is not repaired.

Karl, you’ve been around long enough to remember how this stuff started - people maintaining static hosts.txt files and ftp’ing them across.

People soon realized that didnt work too well (again, mainly for the reasons people keep pointing to, for why you need a unique and centralized set of root servers), and there was the SRI-NIC hosts.txt file that’d get updated centrally and then ftp’d periodically by internet connected hosts.

Then uucp-maps (again centralized, more or less), and then finally a unique set of root servers.

Alternate root proposals are simply trying to re-implement what was originally discarded as unusable for a whole lot of problems other than the scalablity of multiple different sets of hosts files (scalablity would be fixed by using dns instead of hosts.txt, but not all that many of the other issues - which keep getting pointed out to you - would get addressed). So, I for one see absolutely no point in reinventing the wheel based on a long discarded spec.

To Suresh:

The notion of competing roots, much less the forces that might drive a set of nations to establish a competing root system that is in parallal to (and consistent with) the US controlled root, has nothing to do with host files.

Moreover, the zone we are talking about has about 250 entries and, when compressed is all of about 15K bytes, which is a size rather smaller than the typical cutsie icons that are plastered all over web pages.

And since, in the situation we are talking about here, the number of people who would be loading that root zone into servers, would amount to no more than a dozen or so, we are hardly talking about a massive issue nor one that hasnt’ been asked and well answered by todays legacy root server operators.

The issue is not the tiny little zone with 250 entries. Its all the sub zones that branch off from that.  And the necessity of maintaining global uniqueness in domains so that you dont get any number of alternate roots pointing a host somewhere while the “regular” root servers point it elsewhere. I’m interested in finding out why that isn’t considered a problem, and whether that wasn’t one of the major problems with locally maintained hosts.txt files.

Scalablity was already addressed when dns got to be used instead of static hosts.txt files all over the place and it’d be addressed in alternate root nameservers as well.

Uniqueness of a particular dns record, and ensuring that you’d get the same record no matter which resolver in the world you query, on the other hand?  That is the main problem that led to centralization of the hosts.txt file .. and there weren’t all that many hosts back then when SRI-NIC took over maintenance of hosts.txt, or all the while the uucp-maps project was running.

Uniqueness is something you’d find very hard to achieve with an alternate root setup. I dont know if you consider that a bug or a feature.

The GPS system was designed by and is controlled by the DoD and can be used by anyone, free of charge.  So the news reports I’ve heard, which claimed that Galileo would help the EU financially, make no sense.  It will have a VERY hard time making money competing with a free system even if it wil be more accurate.  The funding indicated at http://en.wikipedia.org/wiki/Galileo_positioning_system indicates it was largely pork-motivated, and is FAR from an EU-only project.

The rationale to develop Galileo is that GPS serves military purposes. Formally, GPS can stop correct functioning at the mercy of the US’DoD. European news mention that this did happen in the past. As the US Governament reserves the right to strike military attacks in violation of any convention, civilian prejects relying on the GPS are at risk.

From this point of view, the comparison to the root zone is particularly interesting: Of course the Internet is a vital part for many Nations, including the US. Thus, it can be the target of military attacks, and National Governaments may want to use it for military purposes. Until any single governament controls it, that is.

Regarding consistency of information provided by roots which are technically separate (that is, not synchronized by a single central master), it might be of interest to consider how the telephone system works.

The is no central master “root” file which is the authoritative source.

Information on country codes (such as +41 for Switzerland) is published by ITU.  Information on codes within countries is published by each country.

It is up to each telephone operator to code that information into their switches in such a way that it is consistent and works as intented (that is, when you dial +41 22 730 5887 you reach Richard Hill at ITU in Geneva as opposed to somebody else).

Market forces, not regulation or technology, are what ensures the consistency of telephone numbers around the world.

And there is a good example of one instance in which the lack of forced consistency was a “good thing”.  800 numbers (freephone), were introduced in the USA (as 1-800) many years before the concept was evaluated in other countries.  Since the concept was unknown elsewhere, it was not possible to dial a 1-800 number from outside the USA, which introduced an inconsistency in the global dialling plan.

But market forces (not regulation) resulted in a solution being found.  These days, when you dial a US 1-800 number from outside the USA (at least in the countries where I’ve tried it), you get a recording that says that the call is not toll-free and that you will be billed at international rates.

So we have at least one instance of a system in which the market forces the level of consistency which is required by users, so that neither regulatory nor technical constraints are needed to ensure appropriate consistency.

Richard Hill
Counsellor, ITU-T Study Group 2

Very good points. But those are second or third level tld issues at best.

ITU’s country codes for numbering in each country might be compared to what is served by the roots - 1 for the USA, 91 for India etc. 

I do imagine that there would be at least some mechanism beyond market forces to deal with issues where a telco decides that +1 is not the USA, but say China, or Namibia?

Suresh, you wonder whether there is “some mechanism beyond market forces” to ensure uniform telephone numbering. So far as I am aware, the standard for international telephone numbers, specified by ITU-T E.164 has no legal force. It is merely a convention agreed upon by key participants. Governments of various countries may have the authority to enact laws mandating adherence to this standard, or mandating some other numbering system entirely; I don’t know offhand which telecommunications acts, if any (in whatever country) make such demands. Ultimately the standard is just not worth violating: the potential benefits of doing so are highly dubious, whereas the drawbacks are obvious, so the standard enforces itself.

Analogy between E.164 and DNS breaks down for a couple of reasons. One is that the DNS contains lables representing things other than countries, and these things are quite prominent, as in “.com”. If the DNS did not contain these the similarity would be closer, but as it stands they form a significant difference. The other bone of contention is that ITU-T is an international organisation, whereas ICANN is the creature of the US government. Thirdly, market forces do not command the universal respect of the DNS root quite as surely as they command respect of E.164, mostly because of the non-geographic DNS elements under the unilateral control of the US-based ICANN. John Levine’s recent report shows that market forces are allowing (perhaps even encouraging) root splits in certain cases.

Richard’s mention of 1-800 numbers as an example of good inconsistency is slightly dubious as relates to the DNS. The problem with 1-800 numbers is that they are attached to a special billing arrangement that can only be enforced at a national level (or, more specifically, by special technological measures and cooperation between telephone providers). It’s only a feature that such calls fail to reach their intended destination from international points of origin when the caller is under the false impression that the call will be free.

Please stop the bad analogies everyone.

The GPS is controlled by one military body, the root name servers are controlled by a loose group of organisations (some of them European, indeed the ICANN root had more European based and controlled servers than the ORSN last time I checked) that agree consistency is a good thing. If tomorrow ICANN decided to delete some TLD for political reasons, that consensus would likely break down, and Paul’s support for a backup system might seem farsighted.

Anyone holding up phone numbers as an example of how to ensure global consistency, clearly doesn’t do enough international travelling (did I say get out more?).

Unlike the phone numbering system, the ICANN rooted DNS works exactly the same in every country of the world, and uses mnemonic names instead of long, seemingly random streams of numbers. One is sane, the other a perpetual irritation. As such I expect Skype, or something similar, will soon replace phone numbers with a globally unique identifier, and more functional directory service.

Analogies are helpful for explanation, and education, but they hinder the issue when working to technical solutions (hopefully people working to technical solutions already understand the technologies they are working with), when it is better to state what you mean. Unless of course all one is interested in is persuasion, and politics, rather than engineering.