Domain names and persistence: Report on a workshop

Henry S. Thompson
Jonathan Rees
23 January 2012 (amended 24 Jan)

1. Background

As part of the W3C Technical Architecture Group (TAG)'s long-standing concern with issues around the idea of persistence and the Web, one particular problem on which it seems that progress might be possible is that of the persistence of domain name 'ownership'. As stated in the call for participation:

The vulnerability of any digital material to unexpected or unintended changes in Internet domain name assignment, and hence to the outcome of domain name resolution, is widely recognised. The fact that domain names are not permanently assigned is regularly cited as one of the main reasons why http: URIs cannot be regarded as persistent identifiers over the long term.

With the support of the W3C and the Digital Curation Centre, a workshop was held in Bristol on 8 December 2011 in conjunction with the 7th International Digital Curation Conference.

Virtually all the relevant constituencies were represented by one or more of the attendees, and after a series of short introductory talks, a very wide-ranging and productive discussion ensued. A meeting record, which includes the names of all participants, links to all the presentations and informal minutes of the discussion, is available.

2. Themes

Perhaps not surprisingly given that people chose to come to the workshop, there were reminders throughout the day that the topic of the workshop does matter, that persistence of names on the Web in general does depend on persistance of domain names in particular. The move by the proprietors of several of the larger non-http: URI schemes (e.g. doi: and hdl:) to provide and even promote 'actionable' http: versions of their URIs (via and respectively) was referred to a number of times, and underlines the role of domain names in the overall persistence ecology.

The issue of opacity, that is, of whether names in general, and domain names in particular, should be human-readable, came up repeatedly, although never at great length. There was clearly no consensus on this point, with some arguing that we fool ourselves if we think that URIs are understandable outside a small community of experts, while others thought that the natural-language meaning of the strings which appear in URIs are important both for user confidence and for branding.

Although terminology in this area will always be a challenge, some clarity over key concepts did emerge, in particular as regards the distinction introduced by Jonathan Rees between binding and resolution:

(A means of) creating the relationship between a new name and what it is to name
  • "Binding and ownership are different. Ownership is in principle the right to change binding, but in a persistent system, you won't be changing bindings." JAR
(A process for) looking up a name in order to discover what it names

By construction, for persistent names, binding happens once and is intended to be unique and irrevocable. Resolution, on the other hand, may well be achieved by multiple means.

Broad agreement emerged over the course of the day on the importance of establishing an open community process for gatekeeping access to whatever technical mechanism was eventually established.

The difference between, on the one hand, making it possible to create new domain names which had persistence-enabling properties, and on the other hand making it possible to change the status of some existing domain names to make them more persistable came up repeatedly, with the recognition that doing something for new names might be easier than doing so for existing ones.

Terminology note: we started out talking about these processes in terms of creating "gold-plated" domain names, or of "gold-plating" existing names, but by the end of the day expressed a preference for using the phrase "robust domain names" for what we were trying to create. There's no guarantee that that name will . . . persist :-).

3. Using the .arpa gTLD

One possible approach to domain name persistence was suggested which was new to most, perhaps even all, participants. Gavin Brown of the UK registrars CentralNic pointed out that .arpa has special status already. It is not managed under contract by a registrar, as are all other gTLDs, but rather it's managed by IANA:

The .arpa domain is the “Address and Routing Parameter Area” domain and is designated to be used exclusively for Internet-infrastructure purposes. It is administered by the IANA in cooperation with the Internet technical community under the guidance of the Internet Architecture Board.

Creation of second-level domains under .arpa requires approval of an IETF RFC setting out their intended use. Gavin's original idea was that we could approach the IAB to explore a system of sub-registration for domains under .arpa, e.g., with one new RFC establishing a general mechanism and special status for 'robust' sub-domains of .arpa, plus one further RFC for each such sub-domain. We were hopeful that this approach could be put in place entirely on the basis of agreement with the IAB, not requiring any discussion with or agreement from IANA or ICANN.

This approach would allow the creation of new persistable domain names, but would not in itself change the status of existing ones. Henry Thompson suggested that a further step would be to use the creation of e.g. as a way to simultaneously create a new 'robust' domain name, and to change the status of the existing domain to have the same persistable properties. This further step would require agreement not only from the IAB but also new regulations at the level of ICANN's contracts with some gTLD registrars, at the very least PIR and VeriSign for the .org and .net gTLDs respectively. Opinions differed as to the likelihood of achieving such agreement.

Much work remains to be done, both at the conceptual/technical level with respect to binding and resolution mechanisms, and at the political level with respect to the IAB at least, and ICANN and some registrars in the more complex case of existing domains.

4. Conclusions

No formal resolutions were proposed, but some rough consensus did seem evident to us in the following areas:

Nothing is permanent. We generally preferred to talk about persistence or conditions for persistability.
New gTLD
The possibility of establishing some kind of persistent top-level domain was discussed from time to time, but there was little or no enthusiasm for this. On the one hand the forthcoming opportunity to apply for new generic top-level domain names is not structured in a way likely to be conducive to the kind of governance described above. On a more practical level, we didn't think the necessary agreement among the major consituencies for such an effort could be achieved in time. On the other hand, getting a one-off new gTLD with its own new governance rules through ICANN would involve an enormous amount of political complexity which no-one felt confident could be managed to a successful conclusion.
There was a lot of excitement about the suggestion that the existing gTLD .arpa could be the basis for the creation and management of 'robust' domain names. More work is clearly needed, but this does look like by far the most promising of the ways forward which were discussed at the workshop.
With respect to community review, once it was suggested, we quickly converged on IETF's existing RFC mechanism as at the very least an ideal model for how to do this, and if at all possible the actual way access to 'robust' domains should be managed.

5. Looking to the future

A suitable forum remains to be found for taking the discussion started in Bristol forward. A W3C Incubator Group might [have served, but Incubators are no longer available: the current preferred alternative would be a Community Group]. An obvious thing any future activity needs to do is produce some sort of white paper on the .arpa idea—I've started putting some notes down on a wiki page, part of the overall W3C wiki area for the Domain Names and Persistence topic.

6. Acknowledgements

The organisers wish to record their thanks to all the participants for making this a true workshop, not just a sequence of talking heads. The introductory speakers all responded well to our request to focus on clarifying relevant background and raising issues based on experience. The open discussion was carried out in a spirit of co-operation. As a result real progress was made on a number of fronts, both conceptual and practical.