Observations on the Management of the Internet Address Space

Network Working Group                                          G. Huston
Request for Comments: 1744                                        AARNet
Category: Informational                                    December 1994


                   Observations on the Management of
                       the Internet Address Space

Status of this Memo

   This memo provides information for the Internet community.  This memo
   does not specify an Internet standard of any kind.  Distribution of
   this memo is unlimited.

Abstract

   This memo examines some of the issues associated with the current
   management practices of the Internet IPv4 address space, and examines
   the potential outcomes of these practices as the unallocated address
   pool shrinks in size.  Possible modifications to the management
   practices are examined, and potential outcomes considered.  Some
   general conclusions are drawn, and the relevance of these conclusions
   to the matter of formulation of address management policies for IPv6
   are noted.

1.  Introduction

   The area explicitly examined here is the allocatable globally unique
   IPv4 address space.  Explicitly this includes those address groups
   uniquely assigned from a single comprehensive address pool to
   specific entities which are then at liberty to assign individual
   address values within the address group to individual hosts.  The
   address group is handled by the technology as a single network
   entity.

   At present these addresses are allocated to entities on a freely
   available, first-come, first-served allocation basis, within the
   scope of a number of administrative grounds which attempt to direct
   the allocation process to result in rational use of the space, and
   attempt to achieve a result of a level of equity of availability that
   is expressed in a sense of multi-national "regions" [1].

   In examining the current management policies in further detail it is
   useful to note that the IPv4 address space presents a number of
   attributes in common with other public space resources, and there are
   parallels in an economic analysis of this resource which include:

    - the finite nature of the resource

      This attribute is a consequence of the underlying technology
      which has defined addressed entities in terms of a 32 bit address
      value.  The total pool is composed of 2**32 distinct values (not
      all of which are assignable to end systems).

    - the address space has considerable market value

      This valuation is a consequence of the availability and extensive
      deployment of the underlying Internet technology that allows
      uniquely addressed entities the capability to conduct direct end-
      to-end transactions with peer entities via the Internet.  The
      parameters of this valuation are also influenced by considerations
      of efficiency of use of the allocated space, availability of end
      system based internet technologies, the availability of Internet-
      based service providers and the resultant Internet market size.

    - address space management is a necessary activity

      Management processes are requires to ensure unique allocation and
      fair access to the resource, as well as the activity of continuing
      maintenance of allocation record databases.

   Increasing rates of Internet address allocation in recent years imply
   that the IPv4 address space is now a visibly finite resource, and
   current projections, assuming a continuation of existing demand for
   addresses predict unallocated address space exhaustion in the next 6
   - 12 years (rephrasing current interim projections from the IETF
   Address Lifetime Expectancy Working Group).  There are two derivative
   questions that arise from this prediction.  Firstly what is the
   likely outcome of unallocated address space exhaustion if it does
   occur, and secondly, are there corrective processes that may be
   applied to the current address management mechanisms that could allow
   both more equitable allocation and potentially extend the lifetime of
   the unallocated address space pool.  These two issues are considered
   in the following sections.

2. Outcomes of Unallocated Address Space Exhaustion - No change in
   current Address Management Policies

   As the pool of available addresses for allocation depletes, the
   initial anticipated outcome will be the inability of the available
   address pool to service large block address allocation requests.
   Such requests have already been phrased from various utility
   operators, and the demand for very large address blocks is likely to
   be a continuing feature of address pool management.  It is noted that
   the overall majority of the allocated address space is very
   inefficiently utilised at present (figures of efficiency of use of
   less than 1% are noted in RFC 1466, and higher efficiency utilisation
   is readily achievable using more recent routing technologies, such as
   Variable Length Subnet Masks (VLSM) and disjoint subnet routing).
   Given the continuing depletion of the unallocated address pool, and
   the consequent inability to service all address allocation requests,
   it is a likely outcome of interaction between those entities with
   allocated address space and those seeking address allocation that
   such allocation requests could be satisfied through a private
   transaction. In this situation an entity already in possession of a
   sufficiently large but inefficiently utilised allocated address block
   could resell the block to a third party, and then seek allocation of
   a smaller address block from the remaining unallocated address space.
   The implication is that both address blocks would be more efficiently
   utilised, although it is the entity which has large blocks of
   allocated address space which would be the primary beneficiary of
   such transactions, effectively capitalising on the opportunity cost
   of higher efficiency of address block use.

   Such reselling / trading opportunities which involve the use of the
   unallocated address pool would in all likelihood be a short term
   scenario, as the high returns from this type of trading would
   increase the allocation pressure from the pool and act to increase
   depletion rates as more pressure is placed to claim large address
   blocks for later resale once such blocks are no longer available from
   the unallocated pool.

   Following exhaustion of the unallocated address pool a free trading
   environment in address blocks is a probable outcome, where address
   blocks would be bought and sold between trading entities.  The
   consequent market, if unregulated, would act to price address space
   at a level commensurate with the common expectation of the market
   value of addresses, trading at a price level reflecting both the
   level of demand, the opportunity cost of more efficient address use,
   and the opportunity cost of deployment of additional or alternate
   internetworking technologies to IPv4.  It is interesting to note that
   within such an environment the registry (or whatever takes the place
   of a registry in such an environment) becomes analogous to a title
   office, acting to record the various transactions to ensure the
   continued accuracy of "ownership" and hence acts as a source of
   information to the purchaser to check on the validity of the sale by
   checking on the validity of the "title" of the vendor.  This impacts
   on the characteristic features of Internet address registries, which
   effectively become analogous to "titles offices", which typically are
   structured as service entities with "lodgement fees" used to fund the
   action of recording title changes.  Whether existing registries adapt
   to undertake this new function, or whether other entities provide
   this function is a moot point - either way the function is a
   necessary adjunct to such a trading environment.

   It is also anticipated that in an unregulated environment the trade
   in address blocks would very quickly concentrate to a position of
   address trading between major Internet providers, where a small
   number of entities would control the majority of the traded volume
   (market efficiency considerations would imply that traders with large
   inventories would be more efficient within this trading domain).  It
   is also reasonable to expect that the Internet service providers
   would dominate this trading area, as they have the greatest level of
   vested interest in this market resource.  This would allow the
   Internet service provider to operate with a considerably greater
   degree of confidence in service lifetime expectation, as the service
   provider would be in the position of price setting of the basic
   address resource and be able to generate an address pool as a hedge
   against local address depletion for the provider's client base.
   There is of course the consequent risk of the natural tendency of
   these entities forming a trading cartel, establishing a trading
   monopoly position in this space, setting up a formidable barrier
   against the entry of new service providers in this area of the
   market.  Such a scenario readily admits the position of monopoly-
   based service price setting. Compounding this is the risk that the
   providers set up their own "title office", so that in effect the
   major trading block actually controls the only means of establishing
   legitimacy of "ownership", which in terms of risk of anti-competitive
   trading practices is a very seriously damaged outcome.

   Assuming a relatively low cost of achieving significantly higher
   efficiency address utilisation than at present, then the resultant
   market is bounded only by the costs of agility of renumbering.  Here
   renumbering would be anticipated to occur in response to acquisition
   of a different address block in response to changing local address
   requirements, and the frequency of renumbering may occur in cycles of
   duration between weeks and years.  Markets would also be constrained
   by deployment costs, where local address trading within a provider
   domain would have little cost impact on deployment services (as the
   aggregated routing scenario would be unchanged for the provider and
   the provider's peers) whereas trading in small sized blocks across
   provider domains would result in increased operational service cost
   due to increased routing costs (where efforts to create aggregated
   routing entries are frustrated by the effects of address leakage into
   other routing domains).

   In examining this consequent environment the major technical outcome
   is strong pressure for dynamic host address assignment services,
   where the connection and disconnection of hosts into the Internet
   environment will cause a local state change in allocated addresses
   (which may in turn trigger consequent extended dynamic renumbering
   from time to time to accommodate longer term address usage trends).
   It is also reasonable to predict a strengthening market for dynamic
   address translation technologies, as an alternate client strategy to
   the purchase of large address blocks from the trading market (this
   scenario is the use of a private, potentially non-unique address
   space within the client network, and the dynamic translation of end
   host addresses into a smaller unique Internet routed address pool to
   support external end-to-end sessions), and also the strengthened
   market for firewall boundary technologies which also admit the use of
   private address space within the client domain.

   While it is not possible to accurately predict specific outcomes, it
   would appear to be the case that increasing overall efficiency of
   address utilisation will be most visible only after unallocated
   address pool exhaustion has occurred, as there is then a consequent
   strong economic motivation for such activity across all the entire
   Internet address space.

   As perhaps a cautionary comment regarding evolutionary technologies
   for IPv4, it would also appear to be the case that evolutionary
   technologies will not assume a quantum increase in economic viability
   simply because of unallocated address pool exhaustion.  Such
   technologies will only lever additional advantage over IPv4 once the
   marginal cost of increased IPv4 address space deployment efficiency
   exceeds the marginal cost of deployment of new technologies, a
   situation which may not occur for some considerable time after
   unallocated address pool exhaustion.

3. Modification of Current Internet Address Management Policies

   The three major attributes of the current address allocation
   procedures from the unallocated pool are "first come first served"
   (FCFS) and allocation on a "once and for all" (OAFA) basis, and the
   absence of any charge for address allocation (FREE).

   As noted above, the outcomes of such a process, when constrained by
   the finite quantity of the resource in question, ultimately leads to
   a secondary market in the resource, where initially allocated
   resources are subsequently traded at their market valuation.  This
   secondary trade benefits only those entities who established a
   primary position from the unallocated pool, and it is noted with
   concern that the optimal behaviour while the unallocated pool exists
   is to hoard allocated addresses on the basis that the secondary
   market will come into existence once the pool is exhausted.  Such a
   market does not benefit the original address management operation,
   nor does it necessarily benefit the wider community of current and
   potential interested parties in the Internet community.

   It is also noted that the outcome of a free address allocation policy
   is the vesting of the management of the address space to the larger
   Internet Service Providers, on the basis that in the absence of end
   client address allocation charging policies which have the capability
   of ensuring an independent address management function, those
   entities who have the greatest vested interest in the quality of the
   address allocation and registration function will inevitably fund
   such an operation in the absence of any other mechanism.  The risk
   within this scenario is that placing the major asset of any
   communications medium into the sphere of interest of the current
   entities trading within that medium acts to increase the risk of
   anti-competitive monopolistic trading practices.

   An alternate address management strategy is one of allocation and
   recovery, where the allocation of an address is restricted to a
   defined period, so that the allocation can be regarded as a lease of
   the resource.  In such an environment pricing of the resource is a
   potential tool to achieve an efficient and dynamic address allocation
   mechanism (although it is immediately asserted that pricing alone may
   be insufficient to ensure a fair, equitable and rational outcome of
   address accessibility and subsequent exploitation, and consequently
   pricing and associated allocation policies would be a normative
   approach to such a public resource management issue).

   It is noted that pricing as a component of a public resource
   management framework is a very common practice, where price and
   policy are used together to ensure equitable access, efficient
   utilisation and availability for reallocation after use.  Pricing
   practices which include features of higher cost for larger address
   blocks assist with equitable access to a diversity of entities who
   desire address allocation (in effect a scarcity premium), and pricing
   practices can be devised to encourage provider-based dynamic address
   allocation and reallocation environments.

   In the same fashion as a conventional lease, the leasee would have
   the first option for renewal of the lease at the termination of the
   lease period, allowing the lease to be developed and maintain a
   market value.  Such pricing policies would effectively imply a
   differential cost for deployment of a uniquely addressed host with
   potential full Internet peering and reachability (including local
   reachability) and deployment of a host with a locally defined (and
   potentially non-unique) address and consequent restriction to local
   reachability.

   It is also observed that pricing policies can encourage efficient
   address space utilisation through factors of opportunity cost of
   unused space, balanced by the potential cost of host renumbering
   practices or the cost of deployment of dynamic address allocation or
   translation technologies.

   There are a number of anticipated outcomes of a management mechanism
   which including pricing elements for the IPv4 address space

   Firstly current address space utilisation projections (anticipated
   useful lifetime for the pool of unallocated addresses) would extend
   further into the future due to the factors of cost pressure for more
   efficient address utilisation, and the additional cost of issuing a
   local resource with a globally unique address and the opportunity
   cost of extravagant use of global addresses with purely local
   domains.

   Secondly dynamic host address binding technologies, and dynamic
   network address translation technologies would be anticipated to be
   widely deployed, based on the perceived cost opportunities of using
   such technologies as an alternative to extensive static host address
   binding using globally unique addresses.  Use of such technologies
   would imply further extension of the lifetime of the address pool.

   Such pricing practices could be applied on a basis of all future
   address allocations, leaving those entities with already allocated
   address blocks outside of the lease mechanism.  Alternatively such
   previous allocations could be converted to leases, applying a single
   management policy across the entire address space and accordingly
   levering the maximal benefit from such pricing policies in terms of
   maximising the lifetime of the address space and maximising the value
   of the address space.  In such a situation of conversion some level
   of recognition of previous implicit OAFA allocation policies can be
   offset through delay of conversion to lease and also through
   conversion of such previously allocated addresses to the lease,
   waiving the lease purchase costs in such cases.

4. Internet Environment Considerations

   Pricing for IPv4 addresses as a component of the overall address
   management framework is by no means a novel concept, and despite the
   advantages such pricing policies may offer in terms of outcomes of
   efficiency of utilisation, fair and equitable access, security of
   allocation and consequent market value, and despite the address pool
   exhaustion time offsets such policies offer, it is the undeniable
   case that no explicit pricing policies have been successfully
   introduced into the Internet address allocation processes to date.

   There are two predominate reasons offered in this analysis.  The
   first is the somewhat uncertain nature of the exact origin of primary
   ownership of the IPv4 address space, and the unallocated address pool
   in particular.  The address pool has been administered according to
   policies drafted by the Internet Assigned Numbers Authority (IANA).
   The policies drafted by IANA are effectively policies which are the
   outcome of the same consensus seeking approach used within the
   Internet Standards process, and it is noted that within such an
   environment unilateral declarations of ownership and related
   assertions of policy control have difficulty in asserting an
   effective role within the Internet community and such declarations
   are generally incapable of gathering consensus support (It can be
   argued that "ownership" is not a relevant concept within this domain,
   as the essential attribute of such address elements are their
   uniqueness within the global domain, and such an attribute is only
   feasible through common recognition of a coordinated and reliable
   management environment rather than the historical origin of the
   resource in question).  Secondly there is no formal recognition of
   the address space as being a shared international resource which sits
   within the purview of national public resource management policies
   and administrative entities of each nation, nor is there a
   recognition of the address space as a private resource owned and
   administered by a single entity.

   Recent policy changes, whereby large segments of the unallocated
   address pool have been assigned to international bodies on a regional
   basis, with further assignment to bodies within national contexts,
   have been undertaken with a constant address allocation policy of
   FCFS, OAFA and FREE, and although some effort has been made to
   increase the deployment efficiency through explicit allocation policy
   enumeration, the general characteristics of address allocation are
   unchanged to date (those characteristics being of course FCFS, OAFA
   and FREE).

   One potential scenario is to speculate that pricing processes imposed
   by the address allocation agency are not feasible within the current
   Internet environment to the extent that any such policies could
   significantly motivate increased address deployment efficiency to the
   levels required for longer term unallocated address pool lifetime
   extension.  The lack of capability to employ pricing as a managerial
   mechanism, even to the extent of cost recovery of the allocation and
   subsequent registry maintenance function has a number of possible
   longer term outcomes:

    a) such functions will be restructured and operated from duly
       authorised national administrative bodies for each nation.
       Here the observation that the address pool delegation sequence
       within the current Internet environment has not to date been
       aligned with recognised national public communications resource
       administrative entities is an expression of the major problem
       that the unallocated address pool is not recognised as being
       intrinsically the same public resource entity as the radio
       spectrum or the telephone number space.  The consequence of
       this mismatch between existing public resource management
       structures and IPv4 address space management implies that
       public operation for this activity on a national basis
       is not a commonly observed attribute.  The competency of such
       established public resource management structures in managing
       what continues to be a remarkably vibrant and dynamic
       technology-influenced domain must be questioned.  Potential
       outcomes may possibly include a rational and equitable address
       space management mechanism, but would also in all probability
       include a cost of a heavy damping factor on further
       technological innovation and refinement of the underlying
       technology base upon which the address space is sited as a
       longer term outcome.

    b) such functions are operated (and/or funded) by Internet Service
       Providers.  This is a more common scenario at present in the
       Internet IPv4 environment, and although such an operational
       environment does admit the potential for adequate funding for
       competent administration of the operation, the strong
       association of these entities who have established interests in
       the operation of enterprises based on the provision of services
       across the address space (i.e., strong interest in exploiting
       the address space) has a natural tendency to express domination
       of the market by established interests, threatening fair access
       to the common resource and threatening the open market of
       deployment of the technology.  It is reasonable to suggest that
       such alignments are undesirable from a public policy
       perspective.

    c) such functions are inadequately funded to service the level of
       activity, and / or administrated informally and consequently
       managed poorly, and the essential attribute of reliable address
       space management is not achieved.

   It is noted that these issues are largely unresolved within the
   Internet community today, and tensions between established and
   incoming Internet Service providers over equitable access to the
   unallocated address space pool are a consequent risk.

5. Concluding Observations

   In the absence of the capability to price the management of the
   Internet address space at administrative cost levels, let alone the
   capability to set pricing of address leasing at prices which reflect
   the finite nature of the resource and reflect (even in part) the
   market value of the resource, as a component of overall common
   address management practices, the most likely scenario is a
   continuation of the FCFS, OAFA and FREE address management policies
   until exhaustion of the unallocated address pool occurs.

   It is perhaps a sad reflection of the conflict of short term
   objectives and longer term considerations that the evident short term
   motivations of ready and equitable access to the IPv4 address (which
   were the motivational factors in determining the current Internet
   address allocation policies) run the consequent risk of monopoly-
   based restrictive trade and barrier-based pricing as a longer term
   outcome of unallocated address space exhaustion.

   While free address allocation and the adoption of policies which
   include pricing components both ultimately produce an outcome of
   strong pressure for increased address space utilisation efficiency,
   the removal of the neutral presence of the unallocated address pool
   does induce considerable risk of open market failure within the
   Internet itself if free address allocation policies continue until
   pool exhaustion has occurred.

   Further strengthening of the current FCFS, OAFA and FREE address
   allocation policies, in an effort to induce higher address
   utilization efficiencies across the remaining address space is not a
   viable address management strategy refinement, in so far as the
   trading market will then commence before unallocated pool exhaustion,
   trading in large address blocks which are precluded from such
   strengthened address allocation policies.

   The most negative aspect of this are is that these processes will
   erode levels of confidence in the self regulatory capability of the
   Internet community, such that significant doubts will be expressed by
   the larger community the Internet process is one which is appropriate
   for effective formulation of common administrative policy of one of
   the core common assets of the Internet.

   These outcomes can all be interpreted as policy failure outcomes.

   The seriousness of these outcomes must be assessed in the terms of
   the anticipated timeframe of such policy failure.  Current
   expectations of unallocated address pool lifetime of 6 - 12 years
   does allow the Internet community some time to revisit their methods
   of administrative process definition, but this observation is
   tempered by the IPv6 process and by increasing levels of pressure on
   the address space in terms of growth in address demand through growth
   of deployment of the Internet itself.

   It is perhaps an appropriate conclusion to acknowledge the
   impediments of existing processes to admit any significant process or
   policy change that would produce a more efficient and effective
   address space management regime.

   However it is this policy failure to efficiently utilise the IPv4
   address space through inadequate address pool management policies,
   rather than the exhaustion of the pool per se which is perhaps the
   driving force to design and deploy an evolutionary technology to IPv4
   which possesses as a major attribute a significantly larger address
   space.

   It is also appropriate to conclude that any outside observer of the
   IPv6 refinement process will look to see if there is any evidence of
   experiential learning in address management policies.  If there is to
   be a successor technology for IPv4 it would be reasonable to
   anticipate that associated address pool management mechanisms show a
   greater degree of understanding of public resource space management
   capability in the light of this experience.  If no such evidence is
   forthcoming then there is no clear mechanism to instil sufficient
   levels of consumer and industry confidence in such technologies in
   such a way which would admit large scale public deployment,
   irrespective of the technical attributes of the successor technology.
   Such potential mechanisms may include pricing components irrespective
   of the actual size of the address resource, given that the number's
   uniqueness is a resource with inherent market value irrespective of
   whether scarcity pricing premiums are relevant in such an address
   space.

   It is also appropriate to conclude that continuation of current
   address space management policies run a very strong risk of
   restrictive and monopoly-based trading in address space, with
   consequence of the same trading practices being expressed within the
   deployed Internet itself.

   The immediate action considered to be most appropriately aligned to
   both the interests of the Internet community and the broader public
   community is to examine Internet address space management structures
   which include pricing as well as policy components within the overall
   management mechanism, and to examine the application of such
   mechanisms to both the existing IPv4 address space, and to that of
   any refinement or successor Internet technology base.

6. References

   [1] Gerich, E., "Guidelines for Management of IP Address Space", RFC
       1466, Merit Network, Inc., May 1993.

7. Security Considerations

   Security issues are not discussed in this memo.

8. Author's Address

   Geoff Huston
   Australian Academic and Research Network
   GPO Box 1142
   Canberra  ACT  2601
   Australia

   Phone: +61 6 249 3385
   Fax: +61 6 249 1369
   EMail: Geoff.Huston@aarnet.edu.au