|< Day Day Up >|| |
Once an address space has been defined, there must also be coordinated procedures for handing out unique values within that space and attaching them to users or objects. This process is known as assignment. Assigning unique values to individual users or machines can be viewed as an act of technical coordination. But it can have an economic and policy dimension as well. Figure 2.1 diagrams the relationship. Three distinct criteria that can be applied to the assignment of unique identifiers are represented as distinct layers. The first criterion is the technical coordination that ensures the uniqueness of the assignments. The second layer is economic rationing, that is, the imposition of rules or procedures designed to conserve the resource space. The third layer consists of rules or policies defining or adjudicating rights to names.
Figure 2.1: Three-layer model of assignment
Because of the uniqueness requirement, names and addresses in technological systems are almost always exclusive resources, that is, the assignment of a name or address to one thing necessarily prevents another thing from using the same name or address at the same time. Assignment processes must be organized to maintain this exclusivity. Two or three people cannot be given the same Social Security number without disastrous consequences. Multiple computers on the Internet cannot utilize the same IP address or domain name if they are to communicate reliably with the rest of the Internet. Thus, the assignment process must ensure that the process of giving out addresses or names to users is coordinated to preserve uniqueness and exclusivity.
An identifier space is a finite resource; it can be used up if it is not conserved properly. In addition to preserving the exclusivity of assignments, there may be a need to control the distribution of identifiers to make sure that the resource is not wasted. Are there enough to go around? Should prices or administrative methods be used to ration the resource space? These are important decisions that must be made by an assignment authority (or someone else). Let's call this the economic layer.
In many respects, decisions about economic rationing methods could also be considered policy decisions. Because the size of address spaces is fixed for a long time by standardization decisions that are costly to change, it is not easy to determine what conservation principles to use or how stringently they need to be applied. However, an economic rationing policy deals with a restricted set of issues. Machine-readable identifiers such as IP addresses, credit card numbers, or Ethernet addresses can be thought of as an undifferentiated pool-all the assignment authority needs to worry about is whether the supply of identifiers is sufficient to meet the quantity demanded for the foreseeable future.
As our society has become increasingly information- and communicationsaturated, virtually all the major public network address spaces have had to be expanded. The size of the Ethernet address space (see section 2.5) is being expanded from 48 bits to 64 bits. Internet addresses are (we hope) being expanded from 32 bits to 128 bits. North America altered the syntax of its telephone number plan to make room for many new area codes. [2 ]Since 1996 the toll-free number space in North America has been given four new toll-free codes to keep pace with demand. [3 ]Many countries, including China, have moved to eight-digit local telephone numbers.
Often, the reason for expanding the supply of numbers is not that the available space is fully consumed but that assignment practices delegate large chunks of the space in an inefficient manner. U.S. telephone numbers provide a prime example of inefficient assignment practices. The United States was forced to add 119 new area codes between 1995 and 1999 despite the fact that only 5 percent of the 6.4 billion unique numbers supported by the numbering plan were actually assigned. The problem was that numbers were assigned to the telephone companies' geographic subdivisions in groups with a minimum size of 10,000, even when the areas had only a thousand or so telephone lines. Thus, it is difficult for most assignment authorities to avoid using economic criteria in their practices.
Assignment procedures may be designed to solve policy problems as well as economic and technical problems. If the identifiers are semantically meaningful, an assignment authority may need to make policy decisions about how to resolve competing claims for the same assignment.
The economics of assignment are profoundly affected by who uses the identifier: is it people or machines? As noted before, machine-readable identifiers such as IP addresses, credit card numbers, or Ethernet addresses are an undifferentiated pool. But when people directly interact with identifiers and when the values identifiers take can be meaningful, the market dynamics become far more complex. It is no longer just the quantity of identifiers but their quality that dominates the assignment process.
Think of the difference between two Internet domain names, df5k67tlh.com and music.com. Both are perfectly functional as Web site addresses, but the semantic features of the latter make it far more desirable. People will pay significant sums of money for vanity license plates on their cars. Businesses will sue each other over toll-free telephone numbers that spell words. Households prefer local telephone numbers that are easy to remember. In Hong Kong the Telecommunications Authority holds auctions for local phone numbers that contain lucky numbers. Domain names in the dot-com space based on common words have changed hands for millions of dollars. Semantics can produce huge variations in the economic value of different identifiers in the same space.
Meaning totally subverts the homogeneity of an address space. No two words or symbols mean exactly the same thing. Hence, no two identifiers are perfectly good substitutes for each other in an economic sense. Furthermore, meaning itself varies with the eye of the beholder. The domain name df5k67tlh.com does not seem very valuable, but this assumes that your company's name is not df5k67tlh or that df5-k67-tlh isn't the name of a new wonder drug or a leading rock band. Any apparently meaningless string of characters can become meaningful to some people or acquire secondary meaning through its association with something.
If identifiers are both public and meaningful, legal and policy issues surrounding consumer confusion, fraud, intellectual property, and freedom of speech cannot be avoided. Disputes over who 'deserves' a name or who has a legal right to use it will arise. If you have registered the toll-free telephone number 1-800-COLLECT, and a new toll-free code, 888, is introduced, do you have a 'right' to 1-888-COLLECT or should someone else be allowed to get it? Would the coexistence of these two numbers confuse customers? Similar issues arise in domain names. Is it legitimate for someone who is not French to register france.com or to run a top-level domain .france? Even if we agree that the domain should be limited to the French, how does the assignment authority decide which French organization or person 'deserves' the name?
Or perhaps the technical coordinating body should not be involved in such decisions at all? In the toll-free number space, the U.S. Federal Communications Commission (FCC) decided not to impose trademark protection criteria on the assignments under new toll-free codes, instead leaving such protection to litigation under trademark law. [4 ]
A few technologists have proposed to solve the policy problems created by semantics by eliminating the meaningfulness of the identifiers. For example, proposals to replace meaningful, memorable Internet domain names with meaningless character strings occasionally are put forward in the domain name debate (Vixie 1995). Such 'solutions' are attempts to avoid rather than to cope with the problem. People get involved in business and legal disputes over names because their meaning makes them valuable as identifiers. Eliminating the meaning eliminates the basis for disputes, true, but it also eliminates most of their value. It is like proposing to cure a headache by cutting off one's head.
We have seen how the value of an address assignment can be affected by two economic factors: the scarcity of available unique values and the semantic features of a name or number. I now turn to a third economic factor, almost as important as semantics: the equity a user might have built up in a particular identifier. By equity I mean the investment a user makes in associating her business or organization with a particular public identifier.
Equity, like semantics, is only an issue when the address is part of the human interface. A business's telephone number or Internet domain name may appear on official stationery, business cards, and in directories or Web site links. Equally important, the name or number will be mentally associated with the business or become a part of the personal records of customers and other contacts. This association is economically valuable and tends to accumulate over time. A user who changes or loses an identifier may sacrifice some of that equity or put it at risk. Most of the money put into publicizing an identifier is a sunk cost; it cannot be recovered.
If an identifier is controlled by a service provider, users who want to change service providers will not only risk losing some or all of the equity in the old identifier; they will also have to promote the new address and compensate for temporary confusion and misdirection among their contacts. These are known as switching costs in economics (Shapiro and Varian, 1998). Switching costs may act as a deterrent to competition by making it more difficult for customers to switch service providers. [5 ]
Regulators and policymakers have tried to minimize consumer switching costs by promoting the portability of address assignments across service providers. Various forms of number portability are now being implemented in the telecommunication industry around the world (ITU 1999). Toll-free telephone service in North America was the pioneer of number portability. [6 ]Portability is not an absolute but a quality that is achieved in various degrees. Addresses can be portable across service providers but not across different geographic regions (e.g., you cannot use a North American toll-free number in Europe). Internet domain names have always been portable in the sense that the telecom industry is trying to achieve. That is, the addresses have always been entirely software-based, and assignments have been performed independently of the services provided by infrastructure providers. However, many consumers of Internet services get their domain names from Internet service providers (ISPs) instead of registering them themselves. In those cases, end users are burdened with major switching costs if they attempt to change ISPs. Every time they change their ISP, they must alter their email address, and notify friends and business associates.
How then does an assignment authority distribute identifier resources? The economic techniques that can be used to assign identifiers are the same as those that can be used to ration any resource. The economic literature on this issue is vast, but it is rarely applied specifically to name and address assignment, so it makes sense to recount the techniques here.
First-Come/First-Served One common rationing method is firstcome/first-served: whoever gets there first can grab whatever he likes. That may seem unfair and inefficient, but it has the advantage of extremely low transaction costs. No one has to monitor behavior or enforce any rules (other than the exclusivity requirement, of course). Thus, first-come/ firstserved is a rational way to govern access to abundant, relatively low-value resources, such as parking positions in a suburban shopping mall or domain names back when the Internet was small and noncommercial. Firstcome/first-served is much less problematical when the assignments are homogeneous, that is, when they have no semantic properties. Lawsuits over which organization receives a particular Ethernet identifier are unlikely.
Administrative Fees Administrative fees are another form of rationing. They are charges for identifier assignments imposed on a periodic or onetime basis. The fee amount is basically arbitrary but is used by an assignment authority to discourage those who might consume too much if the assignments were free. The fees may also be used to support the operations of the assignment organization. First-come/first-served methods can be and often are combined with administrative fees.
Market Pricing Market pricing is another common rationing method. Auctions can be used in the initial assignment as a method of resolving contention for resources and to allow the price paid for the assignment to reflect its true scarcity value. A full-fledged market pricing regime goes beyond auctions and allows assignments or entire blocks of the identifier space to be owned and traded. This requires private ownership of parts of the resource space and the freedom of owners to trade those portions in a market. Trading allows the price of the resource to reflect continual variations in supply and demand, thereby creating incentives to use the resource efficiently. Higher (or lower) prices will not only encourage users to find ways to limit (or expand) their consumption but also induce those who might otherwise hoard assignments to release them when the price is right. The transaction costs of creating a market are much higher, but the efficiency characteristics are much better.
Administrative Rules Some assignment authorities will use administrative rules rather than markets to ration scarce number or name assignments. The use of administrative rationing criteria is easier for an assignment authority to implement and more controllable than a market, but it is less able to reflect and adjust to actual supply and demand conditions. As an example, applicants for address block assignments might submit information documenting their 'need' for the assignments, and the assignment authority will evaluate that need. This assessment may be guided by simple administrative rules of thumb or by more complex criteria. At best, administrative rules are a low-transaction cost method of conserving a resource. At worst, they create a growing disconnection between the assignment authority and the actual needs and conditions of users. Some country domain name registrars, for example, imposed a rule that only one domain name should be assigned to an organization. That rule made life easy for the domain administrator but was very frustrating to domain name consumers and completely out of touch with the way domain names have come to be used on the Internet.
Merit Distribution Yet another rationing method is merit distribution. Merit-based assignments occur when the authority in control of the space takes it upon itself to base its assignments upon some extrinsic standard of worthiness. Merit assignment can be considered an extension of the administrative rules method. The authority reviews applicants and decides which ones will best fulfill some policy objective. Procedurally, it is a relatively costly method. It requires extensive documentation to accompany an application for an assignment. Competing, mutually exclusive applications may go through quasi-judicial hearings or be put before the public for comment and criticism. Determinations are more discretionary. The process is often referred to disparagingly as 'beauty contests.' Merit assignments were used by the FCC to assign local broadcasting licenses, and are used by localities to award cable television franchises. Regardless of the efficiency or desirability of merit-based assignment, political reality dictates that it is likely to be used when there are severe constraints on the supply of assignments. If there were only ten telephone numbers to be awarded in the entire world, for example, the process of deciding who got them would be intensely political. Political lobbying and jockeying for influence would almost certainly push the assignment authority into imposing some merit criteria on the awards.
[2 ]The telephone system used to know the difference between an area code and any other part of the number, because area codes took the form NZN, where N is any number from 2 to 9, and Z is either 1 or 0. That syntax restricted the number of available area codes. Now area codes take the form NXX, and one must dial 1 to get into the toll network.
[3 ]The new codes are 888, 877, 866, and 855.
[4 ]FCC CC Docket No. 95-155 Toll Free Service Access Codes, Fourth Report and Order and Memorandum Opinion and Order. Adopted: March 27, 1998. Released: March 31, 1998. Paragraph 7: 'Although we recognize commenters' concerns regarding trademark infringement and unfair competition, we find that those issues properly should be addressed by the courts under the trademark protection and unfair competition laws, rather than by the Commission.'
[5 ]I prefer to speak of 'switching costs' rather than 'lock-in' because the latter is less precise and somewhat judgmental. 'Switching costs' connotes that there are costs associated with change. In some cases these costs are extremely high, in other cases they are not. 'Lock-in' implies that switching is impossible. But businesses incur substantial switching costs all the time. Office headquarters are moved to new addresses, resulting in new phone numbers and new stationery, and generating significant short-term expenses such as confusion and moving expenses. The presence of these switching costs does not necessarily mean that the business is perpetually locked in to a particular landlord. Even something as central to corporate identity as brand names and corporate logos change. Just recently, for example, telephone company giants Bell Atlantic and GTE adopted an entirely new name, Verizon, following their merger. The costs associated with this change probably run in the hundreds of millions of dollars.
[6 ]A 1991 FCC order required the industry to make 800 numbers portable by 1993.
|< Day Day Up >|| |