Status of this Memo This memo defines an Experimental Protocol for the Internet community. Discussion and suggestions for improvement are requested. Please refer to the current edition of the ``IAB Official Protocol Standards'' for the standardization state and status of this protocol. Distribution of this memo is unlimited.
This RFCconsiders X.500 in relation to Internet and UK Domains. A basic model of X.500 providing a higher level and more descriptive naming structure is emphasised. In addition, a mapping of domains onto X.500 is proposed, which gives a range of new management and user facilities over and above those currently available. This specification proposes an experimental new mechanism to access and manage domain information on the Internet and in the UK Academic Community. There is no current intention to provide an operational replacement for DNS.
The Domain (Nameserver) System (DNS) provides a hierarchical resource labelling system [Moc87a] [Moc87b] [Lar83]. Example domains are:
MIT.EDU VENERA.ISI.EDU CS.UCL.AC.UK
Entries usually have a single name, although pointers to entries (not subtrees) may be provided by CNAME records. Information (resource records) is associated with each entry. Name components are typically chosen to be shortish (e.g., ``CS''). RFC 822 mailbox names are closely related [Cro82]. For example:
The local-part of the RFC 822 mailbox can be considered as one level lower in the domain hierarchy.
The OSI Directory, usually known as X.500, provides a very general naming framework [CCI88]. A basic usage of X.500 is to provide Organisationally Structured Names. A Schema for this is defined within the standard. Name components will typically have longish values. This is an example directory name represented in Tabular form:
Country GB Organisation University College London Organisational Unit Computer Science Common Name Stephen E. Hardcastle-Kille
This can also be written in the ``User Friendly Name'' notation defined in [HK91]. This syntax is used for names in the rest of this document:
X.500 has as much relation to the DNS as DNS has to ARP. Paul Mockapetris
This is, essentially, the position adopted here. The basic model is that organisational X.500 is providing a layer of naming at the level above domain names. These structured names can be considered to form a naming layer above domain names. There are the following key differences:
o Organisational X.500 tends to use longer and more descriptive values
o The organisational X.500 DIT is slightly shallower than the DNS tree
o X.500 has a richer information framework than DNS
These differences suggest that the following should NOT be done:
o Represent X.500 information in the DNS
o Have an algorithmic mapping between the two hierarchies
This note proposes to represent DNS information in the DIT, and to provide for a loose coupling between the two trees. This note does not propose an equivalencing of X.500 and Domains.
The proposed model is illustrated in Figure 1. Both an organisational and domain structure is represented in the DIT, by use of appropriate object classes and attribute types. A weak linkage is provided between the two parts of the tree by use of special attributes. Here, the linkage is 1:1, but it may be more complex for some parts of the organisational DIT or domain namespace. The linkage is achieved by use of special attributes, as described in Section 11.
Domains are at the level below X.500 names of the form illustrated in the previous section. However, it is also possible to use X.500 in other ways. In particular, there are benefits from representing Domains in X.500. Note that this is very different to equivalencing, as no attempt is made to represent X.500 information within the domain scheme. There are the following potential advantages:
o Domain Services (DNS and NRS) could be replaced with an OSI service (some may not view this as an advantage). This is particularly attractive for OSI services, where use of a non-OSI directory may be inappropriate.
o For Internet sites, access to domain information (beyond MX records) could be provided for systems registered remotely. For UK Academic Community sites, access to domain information for domains not registered in the NRS could be given. For sites neither on the Internet nor in the UK Academic Community there will usually be even more of an advantage, as they usually have very limited information on domains.
o Assuming that information is downloaded from an X.500 database into a DNS or NRS system, the remote management facilities of X.500 could be used. This is possible because of the extra security features of X.500.
Note: For initial work, the converse situation of information being mastered in Domain Databases and uploaded into the X.500 DIT is more likely.
o User access to the domain data, and in particular searching, could be provided. This would allow users to browse the domain namespace, and to determine information associated with the domains.
o The X.500 framework would allow for additional management information to be stored, and to relate the domain names into a more complex structure of information. For example, this might allow for the managers of a system to be identified, and information on how to contact the manager.
o A facility to map RFC 822 mailbox into a Directory Name (and thus access other user information on the basis of this key) could be provided. This may be useful for the user to determine information about a message originator.
o This technique may be useful to facilitate introduction of security, as it will enable certificates to be associated with domains and mailboxes. This may be very useful for the privacy enchanced mail work [Lin89].
Note that the attribute AssociatedName is defined in Section 11. The manager attribute is defined in the COSINE and Internet naming architecture [BHK91]. It allows a manager to be associated with the domain, which is useful where the manager of the domain is different to the manager of the object defined by the AssociatedName. This will allow any domain to be represented in an X.500 hierarchy. The local part of an RFC 822 mailbox is treated as a special sort of domain component, and so these can be represented in the tree as a natural extension of the hierarchy. For example, consider the mailbox S.Kille@cs.ucl.ac.uk. This will lead to the following structure in the DIT:
Note that the RFC822Mailbox Object Class is a subclass of Domain. Some attributes are allowed to be associated with these objects. There may be other additional management attributes which it is useful to define (e.g., Machine Type, Owner, Location etc.). This allows some information which truly belongs to the domain to be represented there. It also allows for further information to be associated with the domain/mailbox when there is not a relevant part of the organisationally structure DIT to be pointed at. When there is an associated part of the DIT, information from that part of the DIT should not be duplicated in the domain entry.
Wildcards are supported by having "*" as a special domain component name. If there is a need to emulate wildcard matching using the directory, the following algorithm must be employed. For example, the
If A.B.PODUNK.COM is looked up in the directory, the query will fail and indicate that two components are matched. A substitution should be made, and *.*.PODUNK.COM looked up explicitly to identify the associated information.
DNS information can be associated with an entry in the DIT. It is important that this is done in a manner which is exactly equivalent to the information stored in the DNS. This will allow the DIT to have information loaded from the DNS or vice versa. All (authoritative) records associated with a domain will be stored in the DIT. There is no attempt made by the OSI Directory to emulate DNS caching or TTL handling. It is assumed that the master entries are maintained by use of DNS Zone Transfer (or equivalent), and that they can be treated as authoritative. There is a need to define an attribute syntax which represents a DNS record. This then allows DNS records to be stored in the DIT. There are three possible encodings of this record:
ASN.1 Encoded This is the most natural approach in terms of X.500. However, it would require all users of this service to handle the new syntax, which would be awkward. There is a problem with handling the resource format in a general manner.
DNS Binary Encoded Use the formally defined record syntax. This would be convenient for access to the data by DNS related software, but would be an awkward encoding for independent X.500 DUAs.
Text encoded Use of a text encoding derived from the DNS specifications. This is straightforward to map onto DNS protocol, and easy to support in a naive X.500 DUA. This approach is chosen.
The syntax is defined in IA5 characters. The BNF of the record uses the definitions of section 5.1 of RFC 1035. It is
Three examples of this (for domain C.ISI.EDU) might be:
500 A 10.1.0.52 ; Basic address record IN 600 MX 10 VENERA.ISI.EDU. ; MX record 600 IN MX 10 VENERA.ISI.EDU. ; MX record - other order
o The class and TTL may be in either order (following RFC 1035)
o The class defaults to IN
o Domains must always be fully specified (i.e., master file abbreviate rules are not used).
o The TTL for a record must always be present (this saves looking at the parent entry to find the SOA record).
o Records (e.g., SOA) may be multiline. Lines should be separated with the two IA5 characters <CR><LF>.
CNAME records are mapped symmetrically onto Directory Aliases.
This is now defined in terms of attribute and object class definitions. A single record type is defined, as opposed to one attribute type per record type. This allows the definition to not require extension when new DNS Record types are define. However, there is some loss of efficiency if only a single record type is needed, as filtering must be done by the DUA. Similarly, no distinction is made on the basis of DNS class. This means that if there are two class hierarchies, that they must be represented in a single DIT, and that information for different classes must be separated by DUA filtering.
DNSDomain OBJECT-CLASS SUBCLASS OF Domain MAY CONTAIN - DNSRecord "
DNSRecord ATTRIBUTE ATTRIBUTE-SYNTAX IA5String MATCHES FOR EQUALITY
Lookup of a domain is achieved by translating it algorithmically to a Distinguished Name (DN), and reading back attributes desired. This information can be managed and searched in a straightforward fashion.
The information may also be downloaded into a DNS database. This should be done by use of zone transfer. A tool to perform zone transfer (in both directions) between a DNS Server and a DSA would seem to be both straightforward and useful. This would be a key tool in a transition to X.500 based management of the DNS. It would also allow a large part of the DNS namespace to be rapidly made available in an X.500 pilot. Inverse information can be derived by the usual IN-ADDR domain, which will be represented in the same manner in the DIT.
In many cases, Application entities will be closely related to domains. In some cases, it may be appropriate to give Application Entities names which are related to the DNS part of the DIT. In this case, the Domain Name will be used to identify the application, and the entry for the domain will also be of object class Application Process. The children of this entry will identify Application Entities, with names such as ``FTAM Service''.
defines an Object Class for a general network, and shows how it can be subclassed to define technology specific networks.
Network OBJECT-CLASS SUBCLASS OF TOP MAY CONTAIN - Manager, Locality, Description "
IPNetwork OBJECT-CLASS SUBCLASS OF Network MUST CONTAIN -AssociatedDomain"
The Network Object Class allows networks to be defined, and for useful attributes to be associated with the entry. A network will often appear in more than one organisational structure, and this linkage should be achieved by use of aliases. This grouping can facilitate management of networks. The subclass IPNetwork mandates linkage into the DNS part of the DIT. This will be represented in the DIT using the structures of RFC 1101 [Moc89]. Both of the domains which identify the network should be represented in the Object Class. For example, a network might have the (user friendly) name:
UCL-Ethernet, University College London, GB
This would have associated domains 0.0.40.128.IN-ADDR.ARPA and UCL-ETHERNET.UCL.AC.UK. These would both have the analogous DIT representations. For example:
There is a need to associate the organisational X.500 DIT and the DNS tree. The objects represented are different (Domain 6= Organisation; Person 6= RFC 822 Mailbox). Therefore aliasing is not an appropriate linkage. However, in many cases, there is a linkage which is rather
stronger than that implied by the seeAlso attribute. Therefore, we define new attributes, which represent this stronger cross-linkage. The same mechanism can be used to link a domains with an Application Entity or an Application Process. Links from the organisational X.500 DIT to the DNS tree are provided by a new attribute, which could be present in Organisation or Organisational Unit entries.
ObjectWithAssociatedDomain OBJECT-CLASS SUBCLASS OF top MUST CONTAIN -AssociatedDomain"
AssociatedDomain ATTRIBUTE WITH ATTRIBUTE-SYNTAX ia5StringSyntax
For example, the organisational entry:
University College London, GB
would have an attribute:
AssociatedDomain = UCL.AC.UK
Similarly, an RFC 822 mailbox attribute is used to link entries of Person Object Class to their associated DNS entry. This attribute is defined in the Cosine and Internet Naming Architecture [BHK91]. Conversely, there are pointers from the DNS represented tree into the organisational X.500 DIT:
AssociatedName ATTRIBUTE WITH ATTRIBUTE-SYNTAX distinguishedNameSyntax
This attribute is associated with the Domain object class.
This entry is used to provide linkage from the DNS X.500 Hierarchy into the organisational X.500 hierarchy. Where such entries do not exist, attributes in the DNS entry (such as phone number) may be used. It is recommended that information is not duplicated. The preferred setup is for the DNS attributes to be rather skeletal, with pointers into the organisational X.500 DIT.
Experiments should be undertaken to determine the practicality and utility of this scheme, in a pilot environment. A possible approach to this experimentation is described in Appendix A. Object Identifiers have been assigned for this purpose in the Cosine and Internet Naming Architecture [BHK91].
[BHK91] P. Barker and S.E. Hardcastle-Kille. The COSINE and Internet X.500 schema. Request for Comments RFC 1274, Department of Computer Science, University College London, November 1991.
[CCI88] The Directory --- overview of concepts, models and services, December 1988. CCITT X.500 Series Recommendations.
[Cro82] D.H. Crocker. Standard of the format of ARPA internet text messages. Request for Comments 822, University of Delaware, August 1982.
[HK91] S.E. Hardcastle-Kille. Using the OSI directory to achieve user friendly naming. Request for Comments in preparation, Department of Computer Science, University College London, November 1991.
This appendix notes a possible approach to deploying an experiment to evaluate this mechanism. The following components of a possible experiment are noted.
1. User tool. This will take a domain or mailbox as input. This will be looked up in the DIT. This tool should be capable of:
o Attempting to correct user input
o Helping browsing
o Looking up information associated with the domain (or mailbox) and associated name, in particular the manager (of both domain and associated name) and information on the manager (e.g., phone number and mailbox).
o Supply DNS records
o Handle IN-ADDR.ARPA inverse lookups if supplied with an IP Address
o Look up networks
2. A procedural library to allow user interfaces to make easy use of these facilities.
3. Zone transfer tool. This will use the zone transfer protocol to transfer information between a DSA and Domain Nameserver. When writing to the DSA, attributes in an entry which are not DNS records should remain untouched.
4. Linkage patching tool. When the organisational DIT is established, associated domain pointers are usually inserted. A tool can be written to search the DIT and insert the reverse pointers.
5. DNS Manager Tool. This will allow user addition of additional information into the DNS part of the DIT. A standard DUA can probably be used for this.
6. Mailbox download tool. This will allow download of local mailboxes, with pointers to the user entries.
7. Emulation DNS Server, using the Directory as a database. The server should maintain a permanent connection to its local DSA. As there is no OSI bind, the response of this server can be at least as fast as a normal DNS server. There can be two variants of this server.
(a) Using a local DSA as a local database but using DNS distributed operations.
(b) Do all lookups in the directory (using Directory Distributed Operations).
An initial experiment is straightforward. The Zone Transfer Tool (3) can be used to download a large part of the DNS space into a single DSA (there will be some restrictions, as parts of the DNS hierarchy do not permit zone transfer). This can be used repeatedly to maintain the information. The linkage patching tool (4) can be used to put in pointers to parts of the DIT. The user tool can then be used (by all sites participation the the directory pilot) to look up domain information. This will allow the utility of the approach to be evaluated. The manager tool (5) will allow extra information to be added to parts of the DNS tree.
The next stage will be to distribute the DNS part of the DIT over multiple DSAs using Directory distribution techniques. The emulation DNS Server (7) will be useful to ensure that equivalent functionality is being offered by the Directory. It can also be used to examine performance differences. A final step is to master some parts of the DNS hierarchy in the DIT. Because of the zone transfer technique, this will be entirely transparent to the DNS user. Management benefits can then be examined.