Network Working Group
Request for Comments: 2725
Category: Standards Track
C. Villamizar
Avici
C. Alaettinoglu
ISI
D. Meyer
Cisco
S. Murphy
TIS
December 1999

Routing Policy System Security

Status of this Memo

This document specifies an Internet standards track protocol for the Internet community, and requests discussion and suggestions for improvements. Please refer to the current edition of the "Internet Official Protocol Standards" (STD 1) for the standardization state and status of this protocol. Distribution of this memo is unlimited.

Copyright Notice

Copyright © The Internet Society (1999). All Rights Reserved.

Abstract

The RIPE database specifications and RPSL language define languages used as the basis for representing information in a routing policy system. A repository for routing policy system information is known as a routing registry. A routing registry provides a means of exchanging information needed to address many issues of importance to the operation of the Internet. The implementation and deployment of a routing policy system must maintain some degree of integrity to be of any operational use. This document addresses the need to assure integrity of the data by providing an authentication and authorization model.

Table of Contents

   1  Overview  . . . . . . . . . . . . . . . . . . . . . . . . .  3
   2  Background  . . . . . . . . . . . . . . . . . . . . . . . .  3
   3  Implicit Policy Assumptions . . . . . . . . . . . . . . . .  5
   4  Scope of Security Coverage  . . . . . . . . . . . . . . . .  5
   5  Organization of this Document   . . . . . . . . . . . . . .  6
   6  Goals and Requirements  . . . . . . . . . . . . . . . . . .  6
   7  Data Representation . . . . . . . . . . . . . . . . . . . . 10
   8  Authentication Model  . . . . . . . . . . . . . . . . . . . 10
   9  Authorization Model . . . . . . . . . . . . . . . . . . . . 12
     9.1   Maintainer Objects . . . . . . . . . . . . . . . . . . 12
     9.2   as-block and aut-num objects . . . . . . . . . . . . . 13
     9.3   inetnum objects  . . . . . . . . . . . . . . . . . . . 13
     9.4   route objects  . . . . . . . . . . . . . . . . . . . . 14
     9.5   reclaim and no-reclaim attributes  . . . . . . . . . . 14
     9.6   Other Objects  . . . . . . . . . . . . . . . . . . . . 15
     9.7   Objects with AS Hierarchical Names . . . . . . . . . . 16
     9.8   Query Processing . . . . . . . . . . . . . . . . . . . 16
     9.9   Adding to the Database . . . . . . . . . . . . . . . . 17
     9.10  Modifying or Deleting Database Objects . . . . . . . . 19
   10  Data Format Summaries  . . . . . . . . . . . . . . . . . . 20
     10.1  Changes to the RIPE/RPSL Schema  . . . . . . . . . . . 20
   Appendicies
   A  Core and Non-Core Functionality . . . . . . . . . . . . . . 23
   B  Examples  . . . . . . . . . . . . . . . . . . . . . . . . . 23
   C  Technical Discussion  . . . . . . . . . . . . . . . . . . . 26
     C.1   Relaxing requirements for ease of registry   . . . . . 27
     C.2   The address lending issue  . . . . . . . . . . . . . . 28
     C.3   Dealing with non-conformant or questionable older
           data . . . . . . . . . . . . . . . . . . . . . . . . . 29
   D  Common Operational Cases  . . . . . . . . . . . . . . . . . 30
     D.1   simple hierarchical address allocation and route
           allocation . . . . . . . . . . . . . . . . . . . . . . 31
     D.2   aggregation and multihomed more specific routes  . . . 32
     D.3   provider independent addresses and multiple origin
           AS . . . . . . . . . . . . . . . . . . . . . . . . . . 32
     D.4   change in Internet service provider  . . . . . . . . . 32
     D.5   renumbering grace periods  . . . . . . . . . . . . . . 32
   E  Deployment Considerations . . . . . . . . . . . . . . . . . 33
   F  Route Object Authorization Pseudocode . . . . . . . . . . . 35
   Acknowledgments  . . . . . . . . . . . . . . . . . . . . . . . 37
   Intellectual Property Notice . . . . . . . . . . . . . . . . . 38
   References . . . . . . . . . . . . . . . . . . . . . . . . . . 38
   Security Considerations  . . . . . . . . . . . . . . . . . . . 40
   Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . 40
   Full Copyright Statement   . . . . . . . . . . . . . . . . . . 41

1 Overview

The Internet Routing Registry (IRR) has evolved to meet a need for Internet-wide coordination. This need was described in RFC-1787, an informational RFC prepared on behalf of the IAB [14]. The following summary appears in Section 7 of RFC-1787.

While ensuring Internet-wide coordination may be more and more difficult, as the Internet continues to grow, stability and consistency of the Internet-wide routing could significantly benefit if the information about routing requirements of various organizations could be shared across organizational boundaries. Such information could be used in a wide variety of situations ranging from troubleshooting to detecting and eliminating conflicting routing requirements. The scale of the Internet implies that the information should be distributed. Work is currently underway to establish depositories of this information (Routing Registries), as well as to develop tools that analyze, as well as utilize this information.

A routing registry must maintain some degree of integrity to be of any use. The degree of integrity required depends on the usage of the routing policy system.

An initial intended usage of routing policy systems such as the RIPE database had been in an advisory capacity, documenting the intended routing policies for the purpose of debugging. In this role a very weak form of authentication was deemed sufficient.

The IRR is increasingly used for purposes that have a stronger requirement for data integrity and security. This document addresses issues of data integrity and security that is consistent with the usage of the IRR and which avoids compromising data integrity and security even if the IRR is distributed among less trusted repositories.

2 Background

An early routing policy system used in the NSFNET, the policy routing database (PRDB), provided a means of determining who was authorized to announce specific prefixes to the NSFNET backbone. The need for a policy database was recognized as far back as 1989 [6, 4]. By 1991 the database was in place [5]. Authentication was accomplished by requiring confirmation and was a manually intensive process. This solved the problem for the NSFNET, but was oriented toward holding the routing policy of a single organization.

The problem since has become more difficult. New requirements have emerged.

  1. There is a need to represent the routing policies of many organizations.
  1. CIDR and overlapping prefixes and the increasing complexity of routing policies and the needs of aggregation have introduced new requirements.
  1. There is a need to assure integrity of the data and delegate authority for the data representing specifically allocated resources to multiple persons or organizations.
  1. There is a need to assure integrity of the data and distribute the storage of data subsets to multiple repositories.

The RIPE effort specificly focused on the first issue and needs of the European community. Its predecessor, the PRDB, addressed the needs of a single organization, the NSF. The RIPE database formats as described in [2] were the basis of the original IRR.

Routing protocols themselves provide no assurance that the origination of a route is legitimate and can actually reach the stated destination. The nature of CIDR allows more specific prefixes to override less specific prefixes [9, 15, 8]. Even with signed route origination, there is no way to determine if a more specific prefix is legitimate and should override a less specific route announcement without a means of determining who is authorized to announce specific prefixes. Failing to do so places no assurance of integrity of global routing information and leaves an opportunity for a very effective form of denial of service attack.

The Routing Policy System Language (RPSL) [1, 13] was a fairly substantial evolutionary step in the data representation which was largely targeted at addressing the second group of needs. The PRDB accommodated CIDR in 1993 [12] and the RIPE database accommodated the entry of CIDR prefixes from inception, but RPSL provides many needed improvements including explicit support for aggregation.

This document addresses the third group of needs identified above.

While the current implementation supporting weak authentication doesn't guarantee integrity of the data, it does provide extensive mechanisms to make sure that all involved parties get notified when a change is made to the database, whether the change was malicious or intended. This provides inadequate protection against additions. Since the software is increasingly used to configure the major parts of the Internet infrastructure, it is not considered to be adequate anymore to know about and have the ability roll back unintended changes. Therefore, more active security mechanisms need to be developed to prevent such problems before they happen.

A separate document will be needed to address the fourth group of needs.

3 Implicit Policy Assumptions

The authorization model encodes certain policies for allocation of address numbers, AS numbers, and for the announcement of routes. Implicit to the authorization model is a very limited number of policy assumptions.

  1. Address numbers are allocated hierarchically. The IANA delegates portions of the address space to the regional registries (currently ARIN, APNIC and RIPE), which in turn delegate address space to their members, who can assign addresses to their customers.
  1. AS numbers are allocated either singly or in small blocks by registries. Registries are allocated blocks of AS numbers, thereby making the allocation hierarchical.
  1. Routes should only be announced with the consent of the holder of the origin AS number of the announcement and with the consent of the holder of the address space.
  1. AS numbers and IP address registries may be different entities from routing registries.

For subsets of any of these three allocation spaces, network addresses, AS numbers, and routes, these restrictions may be loosened or disabled by specifying a very weak authorization method or an authentication method of "none". However, even when no authentication mechanism is used, all involved parties can be notified about the changes that occurred through use of the existing "notify" attribute.

4 Scope of Security Coverage

This document is intended only to provide an authentication and authorization model to insure the integrity of the policy data in a registry. Only authetication and authorization of additions, deletions, and changes to the database are within the scope of this document. Authentication and authorization of database queries is explicitly out of scope. Mutual authentication of queries, that is authenticating both the origin of the query and the repository from which query results are obtained, is also out of scope.

5 Organization of this Document

Familiarity with RIPE-181 [2] and RPSL [1] is assumed throughout this document. Goals are described in Section 6. Section 7 through Section 9 provide descriptions of the changes and discussion. Section 10 provides a concise summary of data formats and semantics. Appendix C through Appendix E provide additional technical discussion, examples, and deployment considerations.

Goals and Requirements Section 6 provides a more detailed description of the issues and identifies specific problems that need to be solved, some of which require a degree of cooperation in the Internet community.

Data Representation Section 7 provides some characteristics of RPSL and formats for external representations of information.

Authentication Model Section 8 describes current practice, proposes additional authentication methods, and describes the extension mechanism if additional methods are needed in the future.

Authorization Model Section 9 describes the means of determining whether a transaction contains the authorization needed to add, modify, or delete specific data objects, based on stated authentication requirements in related data objects.

Data Format Summaries Section 10 provides a concise reference to the data formats and steps in transaction processing.

Technical Discussion Section C contains some discussion of technical tradeoffs.

Common Operational Cases Section D provides some examples drawn from past operational experience with the IRR.

Deployment Considerations Section E describes some deployment issues and discusses possible means of resolution.

6 Goals and Requirements

The Internet is an open network. This openness and the large scale of the Internet can present operational problems. Technical weaknesses that allow misconfiguration or errant operation in part of the network to propagate globally or which provide potentials for simple denial of service attacks should be eliminated to the extent that it is practical. The integrity of routing information is critical in assuring that traffic goes where it is supposed to.

An accidental misconfiguration can direct traffic toward routers that cannot reach a destination for which they are advertising reachability. This is commonly caused by misconfigured static routes though there are numerous other potential causes. Static routes are often used to provide constant apparent reachability to single homed destinations. Some of the largest ISPs literally have thousands of static routes in their networks. These are often entered manually by operators. Mistyping can divert traffic from a completely unrelated destination to a router with no actual reachability to the advertised destination. This can happen and does happen somewhat regularly. In addition, implementation bugs or severe misconfigurations that result in the loss of BGP AS path information or alteration of prefix length can result in the advertisement of large sets of routes. Though considerably more rare, on a few occasions where this has occurred the results were catastrophic.

Where there is the potential for an accidental misconfiguration in a remote part of the Internet affecting the global Internet there is also the potential for malice. For example, it has been demonstrated by accident that multiple hour outages at a major institution can be caused by a laptop and a dial account if proper precautions are not taken. The dial account need not be with the same provider used by the major institution.

The potential for error is increased by the CIDR preference for more specific routes [8]. If an institution advertises a single route of a given length and a distant router advertises a more specific route covering critical hosts, the more specific route, if accepted at all, is preferred regardless of administrative weighting or any routing protocol attributes.

There is a need to provide some form of checks on whether a route advertisement is valid. Today checks are typically made against the border AS advertising the route. This prevents accepting routes from the set of border AS that could not legitimately advertise the route. Theses checks rely on the use of information registered in the IRR to generate lists of prefixes that could be advertised by a specific border AS. Checks can also be made against the origin AS. If policy information were sufficiently populated, checks could be made against the entire AS path, but this is not yet feasible.

The use of a routing registry can also make it more difficult for prefixes to be used without authorization such as unallocated prefixes or prefixes allocated to another party.

In summary, some of the problems being addressed are:

  • Localizing the impact of accidental misconfiguration made by Internet Providers to that provider's networks only.
  • Eliminating the potential for an Internet provider's customer to use malicious misconfiguration of routing as a denial of service attack if the provider route filters their customers. Localizing the denial of service to that Internet provider only if the immediate Internet service provider does not route filter their customers but other providers route filter the route exchange at the interprovider peering.
  • Eliminating the unauthorized use of address space.

If the data within a routing registry is critical, then the ability to change the data must be controlled. Centralized authorities can provide control but centralization can lead to scaling problems (and is politically distasteful).

Address allocation and name allocation is already delegated. Since delegation can be to outside registries it is at least somewhat distributed [11]. Autonomous System (AS) numbers are allocated by the same authorities. It makes sense to delegate the routing number space in a manner similar to the address allocation and AS number allocation. The need for this delegation of authority to numerous registries increases the difficulty of maintaining the integrity of the body of information as a whole.

As a first step, the database can be somewhat centrally administered with authority granted to many parties to change the information. This is the case with the current IRR. There are a very small number of well trusted repositories and a very large number of parties authorized to make changes. Control must be exercised over who can make changes and what changes they can make. The distinction of who vs what separates authentication from authorization.

  • Authentication is the means to determine who is attempting to make a change.
  • Authorization is the determination of whether a transaction passing a specific authentication check is allowed to perform a given operation.

Different portions of the database will require different methods of authentication. Some applications will require authentication based on strong encryption. In other cases software supporting strong encryption may not be necessary or may not be legally available. For this reason multiple authentication methods must be supported, selected on a per object basis through the specification of authentication methods in the maintainer object "auth" attribute. The authentication methods may range from very weak data integrity checks to cryptographicly strong signatures. The authorization model must sure that the use of weak integrity checks in parts of the database does not compromise the overall integrity of the database.

Additional requirements are placed on the authorization model if the database is widely distributed with delegations made to parties that may not be trustworthy or whose security practices may be lacking. This problem must be addressed in the authorization model in order to enable later evolution to a more distributed routing registry.

Autonomous system numbers can be delegated in blocks and subdelegated as needed and then individual AS numbers assigned. Address allocation is a simple numeric hierarchy. Route allocation is somewhat more complicated. The key attributes in a route object (key with regard to making it unique) contain both an address prefix and an AS number, known as the origin AS. The addition of a route object must be validated against the authorization criteria for both the AS and the address prefix. Route objects may exist for the same prefix with multiple origin AS values due to a common multihoming practice that does not require a unique origin AS. There is often no correlation between the origin AS of a prefix and the origin AS of overlapping more specific prefixes.

There are numerous operational cases that must be accommodated. Some of the more common are listed below. These are explored in greater detail in Appendix D with discussion of technical tradeoffs in Appendix C.

  • simple hierarchical address allocation and route allocation
  • aggregation and multihomed more specific routes
  • provider independent addresses and multiple origin AS
  • changing Internet service providers
  • renumbering grace periods

The authorization model must accommodate a variety of policies regarding the allocation of address space and cannot mandate the use of any one model. There is no standardization of address allocation policies though guidelines do exist [11, 16]. Whether authorization allows the recovery of address space must be selectable on a per object basis and may differ in parts of the database. This issue is discussed further in Appendix C.

7 Data Representation

RPSL provides a complete description of the contents of a routing repository [1]. Many RPSL data objects remain unchanged from the RIPE specifications and RPSL references the RIPE-181 specification as recorded in RFC-1786 [2]. RPSL provides external data representation. Data may be stored differently internal to a routing registry.

Some database object types or database attributes must be added to RPSL to record the delegation of authority and to improve the authentication and authorization mechanisms. These additions are very few and are described in Section 8 and Section 9.

Some form of encapsulation must be used to exchange data. The defacto encapsulation has been the one which the RIPE tools accept, a plain text file or plain text in the body of an RFC-822 formatted mail message with information needed for authentication derived from the mail headers or the body of the message. Merit has slightly modified this using the PGP signed portion of a plain text file or PGP signed portion of the body of a mail message. These very simple forms of encapsulation are suitable for the initial submission of a database transaction.

The encapsulation of registry transaction submissions, registry queries and registry responses and exchanges between registries is outside the scope of this document. The encapsulation of registry transaction submissions and exchanges between registries is outside the scope of this document.

8 Authentication Model

The maintainer objects serve as a container to hold authentication filters. A reference to a maintainer within another object defines authorization to perform operations on the object or on a set of related objects. The maintainer is typically referenced by name in mnt-by attributes of objects. Further details on the use of maintainers are provided in Section 9.1.

The maintainer contains one or more "auth" attributes. Each "auth" attribute begins with a keyword identifying the authentication method followed by the authentication information needed to enforce that method. The PGPKEY method is slightly syntactically different in that the method PGPKEY is a substring.

Authentication methods currently supported include the following. Note that pgp-from is being replaced by the pgpkey (see Section 10 and [18]).

   mail-from  This is a very weak authentication check and is
   
      discouraged.  The authentication information is a regular
      expression over ASCII characters.  The maintainer is authenticated
      if the from or reply-to fields in RFC-822 mail headers are matched
      by this regular expression.  Since mail forgery is quite easy,
      this is a very weak form of authentication.
   
   crypt-pw  This is another weak form of authentication.  The
      authentication information is a fixed encrypted password in UNIX
      crypt format.  The maintainer is authenticated if the transaction
      contains the clear text password of the maintainer.  Since the
      password is in clear text in transactions, it can be captured by
      snooping.  Since the encrypted form of the password is exposed, it
      is subject to password guessing attacks.
   
   pgp-from  This format is being replaced by the "pgpkey" so that the
      public key certificate will be available to remote repositories.
      This is Merit's PGP extension.  The authentication information is
      a signature identity pointing to an external public key ring.  The
      maintainer is authenticated if the transaction (currently PGP
      signed portion of a mail message) is signed by the corresponding
      private key.
   
   pgpkey  This keyword takes the form "PGPKEY-hhhhhhhh", where
      "hhhhhhhh" is the hex representation of the four byte id of the
      PGP public key used for authentication.  The public key
      certificate is stored in a separate object as described in [18].

Repositories may elect to disallow the addition of "auth" attributes specifying weaker forms of authentication and/or disallow their use in local transaction submissions. Repositories are encouraged to disallow the addition of "auth" attributes with the deprecated "pgp- from" method.

Any digital signature technique can in principle be used for authentication. Transactions should be signed using multiple digital signature techniques to allow repositories or mirrors that only use a subset of the techniques to verify at least one of the signatures. The selection of digital signature techniques is not within the scope of this document.

9 Authorization Model

The authorization model must accommodate the requirements outlined in Section 6. A key feature of the authorization model is the recognition that authorization for the addition of certain types of data objects must be derived from related data objects.

With multiple repositories, objects not found in RPSL are needed to control AS delegations and new attributes are needed in existing objects to control subdelegation. The definition of RPSL objects used to implement a distrubuted routing registry system is not within the scope of this document.

9.1 Maintainer Objects

The maintainer objects serve as a container to hold authentication filters. The authentication methods are described in Section 8. The maintainer can be referenced by name in other objects, most notably in the mnt-by attributes of those objects.

Maintainers themselves contain mnt-by attributes. In some cases the mnt-by in a maintainer will reference the maintainer itself. In this case, authorization to modify the maintainer is provided to a (usually very limited) set of identities. A good practice is to create a maintainer containing a long list of identities authorized to make specific types of changes but have the maintainer's mnt-by attribute reference a far more restrictive maintainer more tightly controlling changes to the maintainer object itself.

The mnt-by attribute is mandatory in all objects. Some data already exists without mnt-by attributes. A missing mnt-by attribute is interpreted as the absence of any control over changes. This is highly inadvisable and most repositories will no longer allow this.

An additional maintainer reference can occur through a new attribute, "mnt-routes", and is used in aut-num, inetnum and route objects. The "mnt-routes" attribute is an extension to RPSL and is described in detail in Section 10.

A mnt-routes attribute in an aut-num object allows addition of route objects with that AS number as the origin to the maintainers listed. A mnt-routes attribute in an inetnum object allows addition of route objects with exact matching or more specific prefixes. A mnt-routes attribute in a route object allows addition of route objects with exact matching or more specific prefixes. A mnt-routes attribute does not allow changes to the aut-num, inetnum, or route object where it appears. A mnt-routes may optionally be constrained to only apply to a subset of more specific routes.

Where "mnt-routes" or "mnt-lower" are applicable, any maintainer referenced in the "mnt-by" still apply. The set of applicable maintainers for whatever check is being made is the union of the "mnt-routes" or "mnt-lower" and the "mnt-by". For example, when authorizing a route object software would look at "mnt-routes", if it does not exist, look at "mnt-lower", if that does not exist look at "mnt-by".

9.2 as-block and aut-num objects

An "as-block" object is needed to delegate a range of AS numbers to a given repository. This is needed for authorization and it is needed to avoid having to make an exhaustive search of all repositories to find a specific AS. This search would not be an issue now but would be if a more distributed routing repository is used. Distributed registry issues are not within the scope of this document.

The "as-block" object also makes it possible to separate AS number allocation from registration of AS routing policy.

      as-block:        AS1321 - AS1335

The "aut-num" describes the routing policy for an AS and is critical for router configuration of that AS and for analysis performed by another AS. For the purpose of this document it is sufficient to consider the aut-num solely as a place holder identifying the existence of an AS and providing a means to associate authorization with that AS when adding "route" objects.

The "as-block" object is proposed here solely as a means of recording the delegation of blocks of AS numbers to alternate registries and in doing so providing a means to direct queries and a means to support hierarchical authorization across multiple repositories.

9.3 inetnum objects

The "inetnum" exists to support address allocation. For external number registries, such as those using "[r]whoisd[++]" the "inet-num" can serve as a secondary record that is added when an address allocation is made in the authoritative database. Such records could be added by a address registry such as ARIN as a courtesy to the corresponding routing registry.

      inetnum:        193.0.0.0 - 193.0.0.255
      source:         IANA

9.4 route objects

Currently there are a quite few route objects in more than one registry. Quite a few are registered with an origin AS for which they have never been announced. There is a legitimate reason to be in more than one origin AS.

The "route" object is used to record routes which may appear in the global routing table. Explicit support for aggregation is provided. Route objects exist both for the configuration of routing information filters used to isolate incidents of erroneous route announcements (Section 6) and to support network problem diagnosis.

9.5 reclaim and no-reclaim attributes

A reclaim attribute is needed in as-block, inetnum and route objects. The reclaim attribute allows a control to be retained over more specific AS, IP address or route space by allowing modify and delete privileges regardless of the mnt-by in the object itself.

The reclaim attribute provides the means to enforce address lending. It allows cleanup in cases where entities cease to exist or as a last presort means to correct errors such as parties locking themselves out of access to their own objects. To specify all more specific objects the reclaim attribute value should be "ALL". To allow finer control a set of prefixes can be specified.

A no-reclaim attribute can be used to provide explicit exceptions. A reclaim attribute can only be added to an existing object if the addition of the reclaim attribute does not remove autonomy of existing more specific objects that are covered by the new reclaim attribute.

  1. A reclaim attribute can be added to an existing object if there are no existing exact matches or more specific objects overlapped by the new reclaim attribute, or
  1. if the submitter is listed in the maintainer pointed to by the mnt-by of the objects which are overlapped, or
  1. if any overlapped object is listed in a no-reclaim attribute in the object where the reclaim is being added.

Similarly, a submitter may delete a no-reclaim attribute from an object only when that submitter is the only maintainer listed in the mnt-by attributes of any overlapped objects. If the submitter is not listed in any of the maintainers pointed to by the mnt-by attributes for one or more overlapped object, then the submitter is not permitted to delete the no-reclaim attribute.

If neither a reclaim or no-reclaim attribute is present, then more specific objects of a given object cannot be modified by the maintainer of the less specified object unless the maintainer is also listed as a maintainer in the more specific object. However, the addition of a new route or inetnum object must pass authentication of the largest less specific prefix as part of the authentication check described in Section 9.9.

See Section 10 for a full description of the reclaim and no-reclaim attributes.

9.6 Other Objects

Many of the RPSL ancillary objects have no natural hierarchy the way AS numbers, Internet addresses and routes do have a numeric hierarchy. Some examples are "maintainers", "people" and "role" objects. For these objects, lack of any hierarchy leads to two problems.

  1. There is no hierarchy that can be exploited to direct queries to alternate registries. At some point the query strategy of searching all known registries becomes impractical.
  1. There is no hierarchy on which authorizations of additions can be based.

The first problem can be addressed by considering the name space for each of the ancillary objects to be unique only within the local database and to use explicit references to an external repository where needed. To specify an external repository reference, the object key is preceded by the name of the repository and the delimiter "::". For example a NIC handle may take the form "RIPE::CO19". Currently there is a desire to keep NIC handles unique so the naming convention of appending a dash and the repository name is used. Prepending the repository name provides the unique name space since an object in the RIPE database referencing "CO19" would be interpreted as "RIPE::CO19" by default, but it would still be possible to query or reference "IANA::CO19". There is no possibility of accidentally forgetting to adhere to the conventions when making an addition and the existing objects are accommodated, including cases where name conflicts have already occurred.

The second problem can be partially addressed by using a referral system for the addition of maintainers and requiring that any other object be submitted by a registered maintainer or by IANA. The referral system would allow any existing maintainer to add another maintainer. This can be used in parallel with the addition of other object types to support the maintenance of those objects. For example, when adding a subdomain to the "domain" hierarchy (in the RIPE repository where domains are also handled), even when adding a new domain to a relatively flat domain such as "com", there is already a maintainer for the existing domain. The existing maintainer can add the maintainer that will be needed for the new domain in addition to adding the new domain and giving the new maintainer the right to modify it.

An organization gaining a presence on the Internet for the first time would be given a maintainer. This maintainer may list a small number of very trusted employees that are authorized to modify the maintainer itself. The organization itself can then add another maintainer listing a larger set of employees but listing the more restrictive maintainer in the mnt-by attributes of the maintainers themselves. The organization can then add people and role objects as needed and any other objects as needed and as authorization permits.

9.7 Objects with AS Hierarchical Names

Many RPSL objects do not have a natural hierarchy of their own but allow hierarchical names. Some examples are the object types "as- set" and "route-set". An as-set may have a name corresponding to no naming hierarchy such as "AS-Foo" or it may have a hierarchical name of the form "AS1:AS-Bar".

When a hierarchical name is not used, authorization for objects such as "as-set" and "route-set" correspond to the rules for objects with no hierarchy described in Section 9.6.

If hierarchical names are used, then the addition of an object must be authorized by the aut-num whose key is named by everything to the left of the rightmost colon in the name of the object being added. Authorization is determined by first using the mnt-lower maintainer reference, or if absent, using the mnt-by reference.

9.8 Query Processing

A query may have to span multiple repositories. All queries should be directed toward a local repository which may mirror the root repository and others. Currently each IRR repository mirrors all other repositories. In this way, the query may be answered by the local repository but draw data from others.

The mechanism below when applied to multiple repositories assumes the existence of an attribute for traversal of the repositories. The definition of this attribute is considered a distributed registry issue and is out of scope of this document.

For object types that have a natural hierarchy, such as aut-num, inet-num, and route, the search begins at the root database and follows the hierarchy. For objects types that have no natural hierarchy, such as maintainer, person, and role objects, the search is confined to a default database unless a database is specified. The default database is the same database as an object from which a reference is made if the query is launched through the need to follow a reference. Otherwise the default is generally the local database or a default set by the repository. The default can be specified in the query itself as described in Section 9.7.

In the absense of attributes to traverse multiple registries a search of all repositories is needed. With such attributes the search would proceed as follows. In searching for an AS, the delegation attribute in AS blocks can be consulted, moving the search to data from other repositories. Eventually the AS is either found or the search fails. The search for an inetnum is similar. Less specific inetnums may refer the search to other databases. Eventually the most specific inetnum is found and its status (assigned or not assigned) can be determined. The definition of attributes for traversal of repositories is considered a distrbiuted registry issue and is not within the scope of this document.

The search for a route in the presence of attributes for the traversal of multiple registries is similar except the search may branch to more than one repository. The most specific route in one repository may be more specific than the most specific in another. In looking for a route object it makes sense to return the most specific route that is not more specific than the query requests regardless of which repository that route is in rather than return one route from each repository that contains a less specific overlap.

9.9 Adding to the Database

The mechanism below when applied to multiple repositories assumes the existence of an attribute for traversal of the repositories. The definition of this attribute is considered a distributed registry issue and is out of scope of this document.

The root repository must be initially populated at some epoch with a few entries. An initial maintainer is needed to add more maintainers. The referral-by attribute can be set to refer to itself in this special case (Section 10 describes the referral-by). When adding an inetnum or a route object an existing exact match or a less specific overlap must exist. A route object may be added based on an exact match or a less specific inetnum. The root repository must be initially populated with the allocation of an inetnum covering the prefix 0/0, indicating that some address allocation authority exists. Similarly an initial as-block is needed covering the full AS number range.

When adding an object with no natural hierarchy, the search for an existing object follows the procedure outlined in Section 9.8.

When adding an aut-num (an AS), the same procedure used in a query is used to determine the appropriate repository for the addition and to determine which maintainer applies. The sequence of AS-block objects and repository delegations is followed. If the aut-num does not exist, then the submission must match the authentication specified in the maintainer for the most specific AS-block in order to be added.

The procedure for adding an inetnum is similar. The sequence of inet-num blocks is followed until the most specific is found. The submission must match the authentication specified in the maintainer for the most specific inetnum overlapping the addition.

Adding a route object is somewhat more complicated. The route object submission must satisfy two authentication criteria. It must match the authentication specified in the aut-num and the authentication specified in either a route object or if no applicable route object is found, then an inetnum.

An addition is submitted with an AS number and prefix as its key. If the object already exists, then the submission is treated as a modify (see Section 9.10). If the aut-num does not exist on a route add, then the addition is rejected (see Section C for further discussion of tradeoffs). If the aut-num exists then the submission is checked against the applicable maintainer. A search is then done for the prefix first looking for an exact match. If the search for an exact match fails, a search is made for the longest prefix match that is less specific than the prefix specified. If this search succeeds it will return one or more route objects. The submission must match an applicable maintainer in at least one of these route objects for the addition to succeed. If the search for a route object fails, then a search is performed for an inetnum that exactly matches the prefix or for the most specific inetnum that is less specific than the route object submission. The search for an inetnum should never fail but it may return an unallocated or reserved range. The inetnum status must be "allocated" and the submission must match the maintainer.

Having found the AS and either a route object or inetnum, the authorization is taken from these two objects. The applicable maintainer object is any referenced by the mnt-routes attributes. If one or more mnt-routes attributes are present in an object, the mnt- by attributes are not considered. In the absence of a mnt-routes attribute in a given object, the mnt-by attributes are used for that object. The authentication must match one of the authorizations in each of the two objects.

If the addition of a route object or inetnum contains a reclaim attribute, then any more specific objects of the same type must be examined. The reclaim attribute can only be added if there are no more specific overlaps or if the authentication on the addition is present in the authorization of a less specific object that already has a reclaim attribute covering the prefix range, or if the authentication on the addition is authorized for the modification of all existing more specific prefixes covered by the addition.

9.10 Modifying or Deleting Database Objects

When modifying or deleting any existing object a search for the object is performed as described in Section 9.8. If the submission matches an applicable maintainer for the object, then the operation can proceed. An applicable maintainer for a modification is any maintainer referenced by the mnt-by attribute in the object. For route and inet-num objects an applicable maintainer may be listed in a less specific object with a reclaim attribute.

If the submission is for a route object, a search is done for all less specific route objects and inetnums. If the submission is for an inetnum, a search is done for all less specific inetnums. If the submission fails the authorization in the object itself but matches the reclaim attribute in any of the less specific objects, then the operation can proceed. Section C contains discussion of the rationale behind the use of the reclaim attribute.

A modification to an inetnum object that adds a reclaim attribute or removes a no-reclaim attribute must be checked against all existing inetnums that are more specific. The same check of the reclaim attribute that is made during addition must be made when a reclaim attribute is added by a modification (see Section 9.9).

A deletion is considered a special case of the modify operation. The deleted object may remain in the database with a "deleted" attribute in which case the mnt-by can still be consulted to remove the "deleted" attribute.

10 Data Format Summaries

RIPE-181 [2] and RPSL [1] data is represented externally as ASCII text. Objects consist of a set of attributes. Attributes are name value pairs. A single attribute is represented as a single line with the name followed by a colon followed by whitespace characters (space, tab, or line continuation) and followed by the value. Within a value all whitespace is equivalent to a single space. Line continuation is supported by a backslash at the end of a line or the following line beginning with whitespace. When transferred, externally attributes are generally broken into shorter lines using line continuation though this is not a requirement. An object is externally represented as a series of attributes. Objects are separated by blank lines.

There are about 80 attribute types in the current RIPE schema and about 15 object types. Some of the attributes are mandatory in certain objects. Some attributes may appear multiple times. One or more attributes may form a key. Some attributes or sets of attributes may be required to be unique across all repositories. Some of the attributes may reference a key field in an object type and may be required to be a valid reference. Some attributes may be used in inverse lookups.

A review of the entire RIPE or RPSL schema would be too lengthy to include here. Only the differences in the schema are described.

10.1 Changes to the RIPE/RPSL Schema

One new object type and several attributes are added to the RIPE/RPSL schema. There are significant changes to the rules which determine if the addition of an object is authorized.

The new object type is listed below. The first attribute listed is the key attribute and also serves as the name of the object type.

   as-block        key  mandatory  single    unique
   descr                optional   multiple
   remarks              optional   multiple
   admin-c              mandatory  multiple
   tech-c               mandatory  multiple
   notify               optional   multiple
   mnt-by               mandatory  multiple
   changed              mandatory  multiple
   source               mandatory  single

In the above object type only the key attribute "as-block" is new:

   as-block  This attribute provides the AS number range for an "as-
      block" object.  The format is two AS numbers including the sub-
      string "AS" separated by a "-" delimiter and optional whitespace
      before and after the delimiter.

In order to support stronger authentication, the following keywords are added to the "auth" attribute:

   pgp-from  The remainder of the attribute gives the string identifying
      a PGP identity whose public key is held in an external keyring.
      The use of this method is deprecated in favor of the "pgpkey"
      method.
   
   pgpkey  See [18].

In order to disable authentication and give permission to anyone, the authentication method "none" is added. It has no arguments.

An additional change is the "auth" attribute is allowed to exist in a "person" or "role" object. The "auth" method "role" or "person" can be used to refer to a role or person object and take the "auth" fields from those objects. Care must be taken in implementations to detect circular references and terminate expansion or the references already visited.

A few attributes are added to the schema. These are:

   mnt-routes  The mnt-routes attribute may appear in an aut-num, inet-
      num, or route object.  This attribute references a maintainer
      object which is used in determining authorization for the addition
      of route objects.  After the reference to the maintainer, an
      optional list of prefix ranges (as defined in RPSL) inside of
      curly braces or the keyword "ANY" may follow.  The default, when
      no additional set items are specified is "ANY" or all more
      specifics.  The mnt-routes attribute is optional and multiple.
      See usage details in Section 9.1.
   
   mnt-lower  The mnt-lower attribute may appear in an inetnum, route,
      as-block or aut-num object.  This attribute references a
      maintainer object.  When used in an inetnum or route object the
      effect is the same as a "mnt-routes" but applies only to prefixes
      more specific than the prefix of the object in which it is
      contained.  In an as block object, mnt-lower allows addition of
      more specific as-block objects or aut-num objects.  In an aut-num

object the mnt-lower attribute specifies a maintainer that can be used to add objects with hierarchical names as described in Section 9.7.

   reclaim  The reclaim attribute may appear in as-block, aut-num,
      inet-num, or route objects.  Any object of the same type below in
      the hierarchy may be modified or deleted by the maintainer of the
      object containing a reclaim attribute.  The value of the attribute
      is a set or range of objects of the same type where the syntax of
      the set or range is as defined in RPSL. See Section 9.5 for
      restrictions on adding reclaim attributes.
   
   no-reclaim  The no-reclaim attribute is used with the reclaim
      attribute.  The no-reclaim attribute negates any reclaim attribute
      it overlaps.  See Section 9.5 for restrictions on deleting no-
      reclaim attributes.
   
   referral-by  This attribute is required in the maintainer object.  It
      may never be altered after the addition of the maintainer.  This
      attribute refers to the maintainer that created this maintainer.
      It may be multiple if more than one signature appeared on the
      transaction creating the object.
   
   auth-override  An auth-override attribute can be added, deleted, or
      changed by a transaction submitted by maintainer listed in the
      referral-by.  An auth-override can only be added to a maintainer
      if that maintainer has been inactive for the prior 60 days.  The
      auth-override attribute itself contains only the date when the
      attribute will go into effect which must be at least 60 days from
      the current date unless there is already authorization to modify
      the maintainer.  After the date in the auth-override is reached,
      those identified by the maintainer in the referral-by have
      authorization to modify the maintainer.  This attribute exists as
      a means to clean up should the holder of a maintainer become
      unresponsive and can only take effect if that maintainer does not
      remove the auth-override in response to the automatic notification
      that occurs on changes.

The existing "mnt-by" attribute references the "maintainer" object type. The "mnt-by" attribute is now mandatory in all object types. A new maintainer may be added by any existing maintainer. The "referral-by" attribute is now mandatory in the "maintainer" object to keep a record of which maintainer made the addition and can never be changed. Maintainers cannot be deleted as long as they are referenced by a "referral-by" attribute elsewhere.

A Core and Non-Core Functionality

Most of the objects and attributes described in this document are essential to the authorization framework. These are referred to as being part of the "core" functionality. A few attributes listed here are considered "non-core".

The "reclaim" and "no-reclaim" attributes are a convenience to support flexibility in the implementation of address lending.

The "auth-override" attribute is a convenience to facilitate recovery in an environment where repository data is redistributed in any way.

The "referal-by" attribute is a "core" feature. An individual registry may express its sutonomy by creating a self-referencing maintainer, one whose "referal-by" points to itslef. Other registries can decide on a case by case basis whether to consider such an entry valid. A registry may only allow the "referal-by" to refer to a specific maintainer under the control of the registry. This further restriction is an issue that is purely local to the registry.

B Examples

The examples below leave out some required attributes that are not needed to illustrate the use of the objects and attributes described in this document. Missing are admin-c, tech-c, changed, source. Also missing are attributes such as mnt-nfy, whose use are a good practice but are not strictly required.

To do anything at all a maintainer is needed. At some epoch a a single maintainer is populated in one repository and that maintianer has a referal-by pointing to itself. All others referal-by references can be traced back to that maintainer. At the epoch the as-block AS0- AS65535 and the inetnum 0.0.0.0-255.255.255.255 are also allocated. Other ancilliary object may also be needed to bootstrap.

      mntner:        ROOT-MAINTAINER
      auth:          pgpkey-12345678
      
      mnt-by:        ROOT-MAINTAINER
      referal-by:    ROOT-MAINTAINER

This root maintainer might add a top level maintainer for some organization.

      mntner:        WIZARDS
      descr:         High level Technical Folks
      auth:          pgpkey-23456789
      auth:          pgpkey-3456789a
      mnt-by:        WIZARDS
      referal-by:    ROOT-MAINTAINER

That maintainer might add another who have more limited capabilities.

      mntner:        MORTALS
      descr:         Maintain day to day operations
      auth:          pgpkey-456789ab
      auth:          pgpkey-56789abc
      auth:          pgpkey-6789abcd
      mnt-by:        WIZARDS
      referal-by:    WIZARDS

Note that the WIZARDS can change their own maintainer object and the MORTALS maintainer object but MORTALS cannot.

At some point an as-block is allocated and broken down. In the example below, private number space is used.

      as-block:      AS65500-AS65510
      mnt-by:        SOME-REGISTRY
      mnt-lower:     WIZARDS

Note that a registry has control over the object that they have created representing the allocation, but have given the party to which the allocation was made the ability to create more specific objects. Below this as-block, an aut-num is added. Note that import and export are normally required for a aut-num but are not shown here.

      aut-num:       AS65501
      mnt-by:        WIZARDS
      mnt-lower:     MORTALS

In aut-num above the WIZARDS maintainer can modify the aut-num itself. The MORTALS maintainer can add route objects using this AS as the origin if they also have authorization for the IP number space in a less specific route or inetnum.

We also need an inetnum allocation. In this example the inetnum is allocated to a completely different organization. Again attributes are omited which would normally be needed in an inetnum.

      inetnum:       192.168.144.0-192.168.151.255
      mnt-by:        SOME-REGISTRY
      mnt-lower:     ISP
      reclaim:       ALL

The maintainer ISP can add more specific inetnums or routes with this address space. Note that the registry has declared their ability to reclaim the address space.

If ISP wished to reclaim all allocations but some suballocation of theirs resisted, we might get something like the following in which they will reclaim only the top half of an allocation (possibly if it remains unused).

      inetnum:       192.168.144.0-192.168.147.255
      mnt-by:        ISP
      mnt-lower:     EBG-COM
      reclaim:       192.168.146/23+

If we assume that the maintainer EBG-COM and the maintainer MORTALS want to add a route object, one way to do it is for both parties to sign. If EBG-COM for some reason couldn't aggregate an allocate a single top level route (which is inexcusable these days) or there was a preference for some reason to avoid the joint signature approach on a submission either party could give the other permission to make the addition. A mnt-routes could be added to the aut-num or a mnt-lower could be added to an inetnum.

      aut-num:       AS65501
      mnt-by:        WIZARDS
      mnt-lower:     MORTALS
      mnt-routes:    EBG-COM {192.168.144/23}

With this change to the aut-num the maintainer EBG-COM could add a route with origin AS65501, but only with a limited address range.

      route:         192.168.144/24
      origin:        AS65501
      descr:         These boneheads don't aggregate
      mnt-by:        EBG-COM
      mnt-by:        FICTION::MORTALS

Note that while the maintainer EBG-COM added the object they allowed the maintainer MORTALS the ability to modify it.

If an object ended up in another repository, a single maintainer could still be used. In the example above the notation FICTION::MORTALS indicates that the route object is in a different repository and rather than duplicate the maintainer, a reference is made to the repository in which the MORTALS object resides.

In the example below, a pair of route-sets are added and hierarchical names are used.

      route-set:     AS65501:Customers
      mnt-by:        WIZARDS
      mnt-lower:     MORTALS
      
      route-set:     AS65501:Customers:EBG-COM
      mnt-by:        MORTALS
      mnt-lower:     EBG-COM

Suppose in the 192.168.144/24 object above, only the EBG-COM maintainer is listed. If EBG-COM goes bankrupt, no longer needs address space, and stops responding, it could be difficult to delete this object. The maintainer listed in the EBG-COM referral-by attribute could be contacted. They could add a auth-override attribute to the EBG-COM object. Later they could modify the EBG-COM object and then any objects with EBG-COM in the mnt-by.

      mntner:        EBG-COM
      mnt-by:        EBG-COM
      auth-override: 19990401

The examples above stray significantly from realism. They do provide simple illustrations of the usage of the objects type and attributes described in this document and hopefully in doing some are of some value.

C Technical Discussion

A few design tradeoffs exist. Some of these tradeoffs, the selected solution, and the alternatives are discussed here. Some of the issues are listed below.

  1. Whether to err on the side of permissiveness and weaken authorization controls or risk the possibility of erecting barriers to registering information.
  1. Whether to support enforcible address lending or provide the smaller or end user with ultimate control over the registration of the prefixes they are using.
  1. What to do with older objects that either don't conform to newer requirements regarding minimum authorization, authentication, and accountability, or are of questionable validity.

C.1 Relaxing requirements for ease of registry

If the requirement that an aut-num exists is relaxed, then it is possible for anyone to make use of an unassigned AS number or make use of an assigned AS number for which the aut-num has not been entered. Placing requirements on the entry of aut-num presumes cooperation of the Internet address allocation authority (if separate from the routing registry). The address allocation authority must be willing to field requests to populate skeleton aut-nums from the party for which the allocation has been made. These aut-num must include a reference to a maintainer. A request to the address allocation authority must therefore include a reference to an existing maintainer.

The ability to add route objects is also tied to the existence of less specific route objects or inetnums. The Internet address allocation authority (if separate from the routing registry) must also be willing to field requests to add inetnum records for the party already allocated the address space.

The Internet address allocation authority should also add inetnums and aut-nums for new allocations. In order to do so, a maintainer must exist. If a party is going to connect to the Internet, they can get a maintainer by making a request to the Internet service provider they will be connecting to. Once they have a maintainer they can make a request for address space or an AS number. The maintainer can contain a public key for a cryptographicly strong authorization method or could contain a "crypt-key" or "mail-to" authorization check if that is considered adequate by the registering party. Furthermore an address allocation authority should verify that the request for an AS number or for address space matches the authorization criteria in the maintainer.

Currently only the registries themselves may add maintainers. This becomes a problem for the registry, particularly in verifying public keys. This requirement is relaxed by allowing existing maintainers to add maintainers. Unfortunately the accountability trail does not exist for existing maintainers. The requirement then should be relaxed such that existing maintainers may remain but only existing maintainers that have a "referral-by" attribute can add maintainers. The "referral-by" cannot be modified. This requirement can be relaxed slightly so that a "referral-by" can be added to a maintainer by an existing maintainer with a "referral-by". This will allow the accountability trail to be added to existing maintainers and these maintainers can then add new maintainers.

Verifying that a party is who they claim to be on initial addition, is one of the problems that currently falls upon the AS number and address registry. This problem is reduced by allowing existing maintainers to add maintainers. This may actually make it easier to get maintainers and therefore easier to register. The number authority still must verify that the AS or address space is actually needed by the party making a request.

Authorization checks made during the addition of route objects that refer to AS objects and inetnums strongly rely on the cooperation of the Internet address allocation authorities. The number authorities must register as-blocks, aut-nums, or inetnums as AS numbers or address space is allocated. If only a subset of the number authorities cooperate, then either an inetnum or as-block can be created covering the space that registry allocates and essentially requiring null allocation (for example a "crypt-pw" authentication where the password is given in the remarks in the object or its maintainer) or those obtaining addresses from that number authority will have trouble registering in the routing registry. The authorization model supports either option, though it would be preferable if the number authorities cooperated and the issue never surfaced in practice.

The maintainer requirements can be relaxed slightly for existing maintainers making it easier to register. Relaxing requirements on other objects may defeat the authorization model, hence is not an option.

C.2 The address lending issue

The issue of whether lending contracts should be enforcible is an issue of who should ultimately be able to exercise control over allocations of address space. The routing registry would be wise to stay as neutral as possible with regard to disputes between third parties. The "reclaim" and "no-reclaim" are designed to allow either outcome to the decision as to whether the holder of a less specific inetnum or route object can exercise control over suballocations in the registry. The routing registry itself must decide whether to retain control themselves and if so, should very clearly state under what conditions the registry would intervene. A registry could even go to the extreme of stating that they will intervene in such a dispute only after the dispute has been resolved in court and a court order has been issued.

When an allocation is made by a registry, the registry should keep a "reclaim" attribute in the less specific object and make a strong policy statement that the reclaim privilege will not be used except under very clearly defined special circumstances (which at the very minimum would include a court order). If the allocation is further subdivided the party subdividing the allocation and the party accepting the suballocation must decide whether a "reclaim" can be kept by the holder of the less specific allocation or whether a "no- reclaim" must be added transferring control to the holder of the more specific. The registry is not involved in that decision. Different pairs of third parties may reach different decisions regarding the "reclaim" and any contractual restrictions on its use that may be expressed outside of the registry in the form of a legal contract and ultimately resolved by the courts in the event of a bitter dispute.

By retaining "reclaim" rights the registry retains the ability to abide by a court order. This may only truly become an issue in a distributed registry environment where registries will be rechecking the authorization of transactions made elsewhere and may fail to process the attempt of another registry to abide by a court order by overriding normal authorization to change the registry contents if a reclaim is not present.

C.3 Dealing with non-conformant or questionable older data

Some of the newer requirements include requiring that all objects reference a maintainer object responsible for the integrity of the object and requiring accountability for the creation of maintainers to be recorded in the maintainer objects so that accountability can be traced back from an unresponsive maintainer. In the event that contact information is absent or incorrect from objects and there is any question regarding the validity of the objects, the maintainer can be contacted. If the maintainer is unresponsive, the maintainer that authorized the addition of that maintainer can be contacted to either update the contact information on the maintainer or confirm that the entity no longer exists or is no longer actively using the Internet or the registry.

Many route objects exist for which there are no maintainers and for which inetnum and AS objects do not exist. Some contain the now obsoleted guardian attribute rather than a mnt-by.

It is not practical to unconditionally purge old data that does not have maintainers or does not conform to the authorization hierarchy. New additions must be required to conform to the new requirements (otherwise the requirements are meaningless). New requirements can be phased in by requiring modifications to conform to the new requirements.

A great deal of questionable data exists in the current registry. The requirement that all objects have maintainers and the requirements for improved accountability in the maintainers themselves may make it easier to determine contact information even where the objects are not updated to reflect contact information changes.

It is not unreasonable to require valid contact information on existing data. A great deal of data appears to be unused, such as route objects for which no announcement has been seen in many months or years. An attempt should be made to contact the listed contacts in the object, in the maintainer if there is one, then up the maintainer referral-by chain if there is one, and using the number registry or origin AS contact information if there is no maintainer accountability trail to follow. Experience so far indicates that the vast majority of deletions identified by comparing registered prefixes against route dumps will be positively confirmed (allowing the deletion) or there will be no response due to invalid contact information (in many cases the IRR contact information points to nsfnet-admin@merit.edu).

By allowing the registry to modify (or delete) any objects which are disconnected from the maintainer accountability trail, cleanup can be made possible (though mail header forging could in many cases have the same effect it is preferable to record the fact that the registry itself made the cleanup). Similarly, a mechanism may be needed in the future to allow the maintainer in the referral-by to override maintainer privileges in a referred maintainer if all contacts have become unresponsive for a maintainer. The referral-by maintainer is allowed to add an "auth-override" attribute which becomes usable as an "auth" within 60 days from the time of addition. The maintainer themselves would be notified of the change and could remove the "auth-override" attribute before it becomes effective and inquire as to why it was added and correct whatever problem existed. This can be supported immediately or added later if needed.

D Common Operational Cases

In principle, address allocation and route allocation should be hierarchical with the hierarchy corresponding to the physical topology. In practice, this is often not the case for numerous reasons. The primary reasons are the topology is not strictly tree structured and the topology can change. More specificly:

  1. The Internet topology is not strictly tree structured.
  • At the top level the network more closely resembles a moderately dense mesh.
  • Near the bottom level many attachments to the Internet are multi-homed to more than one Internet provider.
  1. The Internet topology can and does change.
  • Many attachments switch providers to obtain better service or terms.
  • Service providers may modify adjacencies to obtain better transit service or terms.
  • Service providers may disappear completely scattering attachments or they may merge.

Renumbering is viewed as a practical means to maintain a strict numeric hierarchy [16]. It is also acknowledged that renumbering IPv4 networks can be difficult [16, 3, 17]. We examine first the simple case where hierarchy still exists. We then examine the operational cases where either initial topology is not tree structured or cases where topology changes.

D.1 simple hierarchical address allocation and route allocation

This is the simplest case. Large ranges of inetnums are assigned to address registries. These registries in turn assign smaller ranges for direct use or to topologically large entities where allocations according to topology can reduce the amount of routing information needed (promote better route aggregation).

AS objects are allocated as topology dictates the need for additional AS [10]. Route objects can be registered by those with authorization given by the AS and by the address owner. This is never an issue where the maintainer of the AS and the inetnum are the same. Where they differ, either the provider can give permission to add route objects for their AS, or the party allocated the address space can give the provider permission to add route objects for their address space, or both parties can sign the transaction. Permission is provided by adding to maintainer attributes.

D.2 aggregation and multihomed more specific routes

Aggregation is normally not a problem if a provider is aggregating address space allocated to the provider and then suballocated internally and/or to customers. In fact, the provider would be expected to do so. This is not a problem even if the route object for the aggregation is added after the more specific route objects since only less specific objects are considered.

Aggregation is potentially a problem if a provider or a set of providers plan to aggregate address space that was never explicitly allocated as a block to those providers but rather remains the allocation of a address registry. These large aggregations can be expected to be uncommon, but relatively easily dealt with. Superaggregates of this type will generally be formed by topologically close entities who have also managed to draw adjacent address allocations. In effect, the registry must give permission to form such a superaggregate by either giving permission to do so in the mnt-routes of an inetnum or by signing the submission along with the other parties.

D.3 provider independent addresses and multiple origin AS

Provider independent addresses and multihoming arrangement using multiple origin AS present a similar problem to multihoming. The maintainer of the address space and the maintainer of the AS is not the same. Permission can be granted using mnt-routes or multiple signatures can appear on the submission.

D.4 change in Internet service provider

A change in Internet service providers is similar to multihoming. A minor difference is that the AS for the more specific route will be the AS of the new provider rather than the AS of the multihomed customer. Permission can be granted using mnt-routes or multiple signatures can appear on the submission.

D.5 renumbering grace periods

Renumbering grace periods allow a provider who wants to keep an address allocation intact to allow a customer who has chosen to go to another provider to renumber their network gradually and then return the address space after renumbering is completed. The issue of whether to require immediate renumbering or offer renumbering grace periods and how long they should be or whether they should be indefinite has been topic of bitter disputes. The authorization model can support no renumbering grace period, a finite renumbering grace period, or an indefinite renumbering grace period. The "reclaim" attribute described in Section 9.1 provides a means to end the grace period.

E Deployment Considerations

This section describes deployment considerations. The intention is to raise issues and discuss approaches rather than to provide a deployment plan.

The use of routing registries is not yet universally accepted. There still remain Internet providers who see no reason to provide the added assurance of accurate routing information described in Section 6. More accurately, these benefits are viewed as being insufficient to justify the cost. This has been largely caused an inability of a very major router vendor up until recently to handle prefix lists of the size needed to specify routing policy on a per prefix basis.

Another reason cited is that filtering on a prefix basis in an environment where routing registry information is incomplete or inaccurate can interfere with connectivity.

There clearly is a critical mass issue with regard to the use of routing registries. A minority of providers use the existing IRR to filter on a per prefix basis. Another minority of providers do not support the IRR and generally fail to register prefixes until connectivity problems are reported. The majority of providers register prefixes but do not implement strict prefix filtering.

Deploying new authentication mechanisms has no adverse consequences. This has been proven with Merit's deployment of PGP.

In deploying new authorization mechanisms, a major issue is dealing with existing data of very questionable origin. A very large number of route objects refer to prefixes that have not been announced for many years. Other route objects refer to prefixes that are no longer announced with the origin AS that they are registered with (some were incorrectly registered to start with). There are many causes for this.

  1. During the transition from the NSFNET PRDB to the RADB a large number of prefixes were registered with an origin AS corresponding to the border AS at which the NSFNET had once heard the route announcements. The PRDB did not support origin AS, so border AS was used. Many of these routes were no longer in use at the time and are now routed with a submitter listed as "nsfnet- admin@merit.edu".
  1. As CIDR was deployed, aggregates replaced previously separately announced more specific prefixes. The route objects for the more specific prefixes were never withdrawn from the routing registries.
  1. Some prefixes are simply no longer in use. Some networks have been renumbered. Some network no longer exist. Often the routing registry information is not withdrawn.
  1. As provider AS adjacencies changed and as end customers switched providers often the actual origin AS changed. This was often not reflected by a change in the routing registry.

Inaccuracies will continue to occur due to the reasons above, except the first. The hierarchical authorization provides greater accountability. In the event that the contacts for specific objects become unresponsive traversal up the authorization hierarchy should help identify the parties having previous provided authorization. These contacts may still have sufficient authorization to perform the necessary cleanup. This issue is discussed in Section C.

A great deal of information is currently missing in the IRR. Quite a few AS have no aut-num. Quite a lot of data has no maintainer and the vast majority of maintainers use only the weakest of authentication methods. Very little can be done by the registries to correct this. The defaults in the cases of missing objects needed for authorization has to be to make no authentication checks at all.

The transition can be staged as follows:

  1. Add and make use of stronger authorization models.
  1. Make schema modifications necessary to support delegations.

3. Add delegation attributes needed for query traversal.
4. Base query traversal on delegations rather than a search of all

known registries.

  1. Obtain the cooperation of the address registries for the purpose of populating the "inetnum" entries on an ongoing basis.
  1. Add hierarchical authorization support for critical object types, "aut-num", "inetnum" and "route".
  1. Add the requirement that database object either be in use or have valid contact information and if queries are made by the registry a response from a contact indicating that the object serves a purpose if it is not clear what its use is.
  1. Begin to purge data which is clearly not in use and for which there is no valid contact information or no response from the contacts.

Deployment of hierarchical authorization requires cooperation among the existing routing registries. New code will have to be deployed. In some cases minimal development resources are available and substantial inertia exists due to the reliance on the current repository and the need to avoid disruption.

If hierarchical authorization of route objects depends on the existence of address registration information, minimal cooperation of the currently separate address registries is required. The extent of the cooperation amounts to sending cryptographically signed transactions from the address registry to the number registry as address allocations are made or providing equivalent access to new address allocations.

Currently most registries return query results from all of the known repositories using their mirrored copies. Cross registry authorizations are not yet implemented. Minimal schema changes have to be made to support the ability to delegate objects for which there is an authorization hierarchy and to support queries and references to other repositories. In the case of AS delegations, "as-block" need to be created solely for the purpose of traversal.

F Route Object Authorization Pseudocode

The following list provides a brief review of basic concepts.

  1. The route object submission must satisfy two authentication criteria. It must match the authentication specified in the aut- num and the authentication specified in either a route object or if no applicable route object is found, then an inetnum.
  1. When checking for prefix authorization, an exact route object prefix match is checked for first. If there is not an exact match then a longest prefix match that is less specific than the prefix is searched for. If the route prefix search fails, then a search is performed for an inetnum that exactly matches the prefix or for the most specific inetnum that is less specific than the route object submission.

The search for an inetnum should never fail but it may return an unallocated or reserved range. The inetnum status must be "allocated" and the submission must pass it's maintainer authorization in order to get authorization from an inetnum. So an unallocated or reserved range inetnum will cause the route object submission to fail.

  1. A route object must pass authorization from both the referenced aut-num object and the route or inetnum object. Authorization shall be tested using the maintainer(s) referenced in the "mnt- routes" attribute(s) first. If that check fails, the "mnt-lower" attributes are checked. If that check fails the "mnt-by" attributes are used for the authorization check.
  1. The "reclaim" attribute can appear in inetnum, route and as-block objects and provides a means to support address lending. "reclaim" gives authorization over more specific objects, regardless of the "mnt-by" in the object. The value of a "reclaim" attribute can be a list or set of objects to provide finer grain control.

The "reclaim" attribute is important to this discussion since it affects prefix/origin authentication when a new route object is submitted.

The "no-reclaim" attribute is used to provide explicit exceptions.

The following pseudocode outlines the algorithm used to check for proper authorization of a route object submission.

Case #1. Route object add

(ie, no exact prefix/origin match exists).

    /* first check the aut-num authorization */

if ( the referenced aut-num object does not exist or

         the aut-num authorization fails )
      authorization fails
    
    /* next we check for prefix authorization */

if ( a less specific route(s) with the longest prefix is found ) [

if ( authorization does not pass for at least one of the less

           specific route(s) )
        authorization fails
    
    /* now check for a "reclaim" attr */

if ( the object has a "reclaim" attribute ) [

        if ( no more-specifics exist
             OR a less specific exists which passes
                authorization and has a "reclaim" attribute
    
             OR all more specifics routess pass modify authorization )
          authorization passes
        else
          authorization fails
      ] else
        authorization passes
    ]
    
    /* there are no less specific routes to check for prefix
       authentication, so we need to try and get authorization from an
       inetnum object */

if ( ( an inetnum is found that is an exact match

           OR is less specific and it's status is "allocated" )
         AND a maintainer referenced by the inetnum
             passes authorization )
      authorization succeeds
    
    /* if there is no inetnum or route object then then
       authorization fails.  This should never happen if
       the DB is initialized properly. */

authorization fails.

    Case #2.  Route object modify/delete
              (ie, exact prefix/origin match exists).

if ( the mnt-by passes authorization )

      authorization succeeds
    
    /* if the authorization did not pass from the matched object,
       we can still get authorization from a less specific route if it
       has a "reclaim" attribute and we pass authorization */
    
    if ( a less specific route or inetnum object passes authorization
         AND has a "reclaim" attribute applicable to
             the object to be modified )
      authorization succeeds
    else
      authorization fails

Acknowledgments

This document draws ideas from numerous discussions and contributions of the IETF Routing Policy System Work Group and RIPE Routing Work Group. Earlier drafts of this document listed Carol Orange as a co- author. Carol Orange made contributions to this document while at RIPE.

Gerald Winters provided the pseudocode in an email message to the RIPE dbsec mailing list that was the basis of the pseudocode found in appendix F. Susan Harris provided comments and numerous editorial corrections.

Intellectual Property Notice

The IETF takes no position regarding the validity or scope of any intellectual property or other rights that might be claimed to pertain to the implementation or use of the technology described in this document or the extent to which any license under such rights might or might not be available; neither does it represent that it has made any effort to identify any such rights. Information on the IETF's procedures with respect to rights in standards-track and standards-related documentation can be found in BCP-11. Copies of claims of rights made available for publication and any assurances of licenses to be made available, or the result of an attempt made to obtain a general license or permission for the use of such proprietary rights by implementors or users of this specification can be obtained from the IETF Secretariat.

The IETF invites any interested party to bring to its attention any copyrights, patents or patent applications, or other proprietary rights which may cover technology that may be required to practice this standard. Please address the information to the IETF Executive Director.

References

    [1]  Alaettinoglu, C., Bates, T., Gerich, E., Karrenberg, D., Meyer,
         D., Terpstra M. and C. Villamizar, "Routing Policy
         Specification Language (RPSL)", RFC 2280, January 1998.
    
    [2]  Bates, T., Gerich, E., Joncheray, L., Jouanigot, J-M.,
         Karrenberg, D., Terpstra, M. and J. Yu, "Representation of IP
         Routing Policies in a Routing Registry (ripe-81++)", RFC 1786,
         March 1995.
    
    [3]  Berkowitz, H., "Router Renumbering Guide", RFC 2072, January
         1997.
    
    [4]  Braun, H-W., "Models of policy based routing", RFC 1104, June
         1989.
    
    [5]  Braun, H-W. and Y. Rekhter, "Advancing the NSFNET routing
         architecture", RFC 1222, May 1991.
    
    [6]  Clark, D., "Policy routing in Internet protocols", RFC 1102,
         May 1989.
    
    [7]  Crocker, D., "Standard for the format of ARPA Internet text
         messages", STD 11, RFC 822, August 1982.
    
    [8]  Fuller, V., Li, T., Yu, J. and K. Varadhan, "Classless Inter-
         Domain Routing (CIDR): an Address Assignment and Aggregation
         Strategy", RFC 1519, September 1993.
    
    [9]  Internet Engineering Steering Group and R. Hinden,
         "Applicability Statement for the Implementation of Classless
         Inter-Domain Routing (CIDR)", RFC 1517, September 1993.
   
   [10]  Hawkinson, J. and T. Bates, "Guidelines for creation,
         selection, and registration of an Autonomous System (AS)", RFC
         1930, March 1996.
   
   [11]  Hubbard, K., Kosters, M., Conrad, D., Karrenberg, D. and J.
         Postel, "Internet Registry IP Allocation Guidelines", BCP 12,
         RFC 2050, November 1996.
   
   [12]  Knopper, M.  and S. Richardson, "Aggregation Support in the
         NSFNET Policy-Based Routing Database", RFC 1482, June 1993.
   
   [13]  Meyer, D., Prior, M., Alaettinoglu, C., Schmitz, J. and Carol
         Orange, "Using RPSL in Practice", RFC 2650, August 1999.
   
   [14]  Rekhter, Y., "Routing in a Multi-provider Internet", RFC 1787,
         April 1995.
   
   [15]  Rekhter Y. and T. Li, "An Architecture for IP Address
         Allocation with CIDR", RFC 1518, September 1993.
   
   [16]  Rekhter Y. and T. Li, "Implications of Various Address
         Allocation Policies for Internet Routing", RFC 2008, October
         1996.
   
   [17]  Rekhter, Y., Lothberg, P., Hinden, R., Deering, S. and J.
         Postel, "An IPv6 Provider-Based Unicast Address Format", RFC
         2073, January 1997.
   
   [18]  Zsako, J., "PGP Authentication for RIPE Database Updates", RFC
         2726, December 1999.

Security Considerations

This document primarily addresses authorization rules for making additions, deletions, and changes to routing policy information repositories. The authentication of these transactions through strong cryptographic means are addressed by [18], referenced thorughout this document. The authorization rules are designed such that the integrity of any transaction can be verified independently by any party mirroring a repository to insure that rules are adhered to. To accomplish this the mirror must contain data already known to be properly authorized. In other words, the mirror must be complete and authentication and authorization checks must be made continuously as changes to the repository are recieved and processed in order.

Authentication alone does not provide a complete security model. Current practice specifies authorization for deletions and changes only, not for additions. The authorization rules provide here complete the security model for additions, deletions, and changes by very explicitly defining rules for addition and clarifying procedures for handling exception cases such as organizations which have ceased to exist and therefore become entirely unresponsive.

Authentication and authorization of queries is explicitly stated to be out of scope of this document.

Authors' Addresses

   Curtis Villamizar
   Avici Systems
   EMail: curtis@avici.com
   
   Cengiz Alaettinoglu
   ISI
   EMail: cengiz@ISI.EDU
   
   David M. Meyer
   Cisco
   EMail: dmm@cisco.com
   
   Sandy Murphy
   Trusted Information Systems
   EMail: sandy@tis.com

Full Copyright Statement

Copyright © The Internet Society (1999). All Rights Reserved.

This document and translations of it may be copied and furnished to others, and derivative works that comment on or otherwise explain it or assist in its implementation may be prepared, copied, published and distributed, in whole or in part, without restriction of any kind, provided that the above copyright notice and this paragraph are included on all such copies and derivative works. However, this document itself may not be modified in any way, such as by removing the copyright notice or references to the Internet Society or other Internet organizations, except as needed for the purpose of developing Internet standards in which case the procedures for copyrights defined in the Internet Standards process must be followed, or as required to translate it into languages other than English.

The limited permissions granted above are perpetual and will not be revoked by the Internet Society or its successors or assigns.

This document and the information contained herein is provided on an "AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.

Acknowledgement

Funding for the RFC Editor function is currently provided by the Internet Society.