Network Working Group L. Daigle Request for Comments: 2967 Thinking Cat Enterprises Category: Informational R. Hedberg Catalogix October 2000
TISDAG - Technical Infrastructure for Swedish Directory Access Gateways
Status of this Memo
This memo provides information for the Internet community. It does not specify an Internet standard of any kind. Distribution of this memo is unlimited.
Copyright Notice
Copyright (C) The Internet Society (2000). All Rights Reserved.
Abstract
The strength of the TISDAG (Technical Infrastructure for Swedish Directory Access Gateways) project's DAG proposal is that it defines the necessary technical infrastructure to provide a single-access- point service for information on Swedish Internet users. The resulting service will provide uniform access for all information -- the same level of access to information (7x24 service), and the same information made available, irrespective of the service provider responsible for maintaining that information, their directory service protocols, or the end-user's client access protocol.
The overarching goal of this project is to develop the necessary technical infrastructure to provide a single-access-point service for searching for whitepages information on Swedish Internet users. The service must be uniform for all information -- the same level of access to information (7x24 service), and the same whitepages information made available, irrespective of the service provider responsible for maintaining that information.
The strength of the TISDAG project's DAG proposal is that it defines the necessary technical infrastructure to provide a single-access- point service for information on Swedish Internet users. The resulting service will provide uniform access for all information --
Daigle & Hedberg Informational [Page 5]
RFC 2967 TISDAG October 2000
the same level of access to information (7x24 service), and the same information made available, irrespective of the service provider responsible for maintaining that information, their directory service protocols, or the end-user's client access protocol.
Instead of requiring centralized mirroring of complete information records from Swedish directory service providers, the DAG system uses a well-defined index object summary of that data, updated at the directory service provider's convenience. When an end-user queries the DAG, the referral information is used (by the end-user's software, or by a module within the DAG, as appropriate) to complete the final query directly at the directory service provider's system. This ensures that the end-user gets the most up-to-date complete information, and promotes the directory service provider's main interest: its service. The architecture of the DAG itself is very modular; support for future protocols can be added in the operational system.
Requirements: As a service, the DAG system will have several different types of users. In order to be successful, those users' needs (requirements) must be met. This in turn defines certain constraints, or system requirements, that must be met. This section aims to capture the baseline requirement assumptions to be addressed by the system, and thus lays the groundwork on which the rest of the proposed system is built.
Functional Specification Overview: Working from the users' requirements, specific technologies and functionality details are outlined to architect a system that will meet the stated requirements. This includes a conceptual architecture for the system. While the Requirements section outlines the needs the different users have for the eventual DAG system, implementing and providing the eventual service will entail constraints or conditions that need to be met in order to be able to participate in the overall system.
Architecture: Once the system has been defined conceptually, a proposed software architecture is specified to produce the desired functionality and meet the stated requirements.
Software Specifications: This section provides the specifications for software components to meet the architecture described above.
Daigle & Hedberg Informational [Page 6]
RFC 2967 TISDAG October 2000
Service Specifications: Once the software has been designed, the success of the DAG system will rest on its operational characteristics. Details of service requirements are given in this section.
DAG-CAP: Client Access Point -- point of communication between client-access software and the DAG system.
DAG-System: The Directory Access Gateway system resulting from the TISDAG project. A collection of infrastructural software and services for the purpose of providing unified access to Swedish whitepages information.
DAG/IP: DAG-Internal Protocol -- communication protocol used between software components of the DAG.
End-User: People performing White Pages searches and look-ups (via various forms of client software).
DAG-SAP: Service Access Point -- point of communication between the DAG and WDSP software.
WDSP: Whitepages Directory Service Provider -- ISPs, companies, or other interested entities.
Whitepages Information: Collected information coordinates for individual people. This typically includes (but is not limited to) a person's name, and e-mail address.
There are 2 primary classes of users for the proposed Whitepages directory access gateway:
- End-users - WDSPs
As outlined below, needs of each of these user classes imposes a set of constraints on the design of the DAG system itself. Some of the requirements shown below are assumed starting criteria for the DAG service; others have been derived from data collected in the Technical Survey or other expertise input.
The End-User is to be provided with a specific set of search types:
Name Name + Organization Role + Organization Name + Locality Name + Organization + Locality Role + Organization + Locality
The search results will, if available, include the following information for each "hit":
- Full name - E-mail address - Role - Organization - Locality - Full address - Telephone numbers
Access to the service must be available through reasonable and current protocols -- such that directory-service-aware software can make use of it seamlessly, and there are no reasonable technological impediments to making this service useful to all Swedish Internet users.
Following on that, its responses are expected to be timely; a standard search should not take more time than the average access to a web-server.
Given that the WDSPs that participate in this service are already in the business of providing a service of whitepages information, they have certain requirements that must be respected in order to make this a successful and useful service to all concerned.
The DAG system must provide reasonable assurances of data integrity for WDSPs; the information the End-User sees should correspond directly to that provided by the WDSPs. The DAG system should be non-preferential in providing whitepages information -- the service is to the End-User, and the source of whitepages information should not influence the search and information presentation processes.
Daigle & Hedberg Informational [Page 8]
RFC 2967 TISDAG October 2000
The DAG system must be able to reflect information updates within a reasonable time after receipt from WDSPs; on the flip side, while the DAG system will function best with regular updates from WDSPs, the update and participation overhead for WDSPs should be held within reasonable bounds of what the WDSP should do to support regular access to its information.
Furthermore, given that WDSPs provide directory service information with an eye to value-added service, wherever possible End-Users should be redirected to the WDSP responsible for individual directory service entries for final and further information.
In the TISDAG pilotservice we have decided to apply some limitations as to what is specified for the DAG/IP. These limitations are presented in this text in the following manner:
The conceptual environment of the DAG system can be described in three major components:
- client access software for end-users - the DAG system core - WDSP directory service software
Daigle & Hedberg Informational [Page 9]
RFC 2967 TISDAG October 2000
This is illustrated in Figure 3.1
The DAG (Directory Access Gateway) is the infrastructural core of the service; it maintains the necessary data and transformation facilities to permit the smooth connection of diverse directory service Client Software to the existing WDSPs' directory servers. The key challenges in designing this portion of the system are:
Quantity of data -- the quantity of whitepages information that will be made available, and diversity of its sources (different WDSPs) introduce challenges in terms of finding a structure that will allow efficient searching, and facilitate the timeliness of updating the necessary information.
Multiplicity of access protocols -- in order to support the use of existing whitepages-aware software with a minimum of perturbation, the DAG system will have to present a uniform face in several different access protocols, each with its own information search and representation paradigm.
This specification will outline the following areas:
- the functioning of the DAG core itself - the interface between the DAG core and End-Users' Directory Service Access software - the interface between the DAG core and Directory Services Servers
In order to reduce the quantity of data the DAG itself must maintain, and to keep the maintenance of the whitepages information as close as possible to the source of information (the WDSPs themselves), the DAG will only maintain index information and will use "query routing" to efficiently refer End-User queries to WDSPs for search refinement and retrieval of information. Although originally developed for the Whois++ protocol, query routing is being pursued in a protocol- independent fashion in the IETF's FIND WG, so the choice of this approach does not limit the selection and support of whitepages access protocols.
The DAG will look after pursuing queries for access protocols that do not support referral mechanisms. In order to achieve the support of multiple access protocols and differing data paradigms, the DAG will be geared to specifically support a limited set of whitepages queries.
The DAG will provide responses including the agreed-upon data. For access protocols that can handle referrals, responses will be data and/or referrals in that query protocol. These are Whois++ and LDAPv3. N.B.: the LDAPv3 proposal defines a referral as a URL; no limitation is placed on the access protocol. However it cannot be assumed that all clients will be able to handle all access protocols, so only referrals to LDAPv3 servers will be returned.
- Searchable Attributes - Matching semantics - Character sets
These, in conjunction with the DAG schema, defined in Appendix A, form the basis of the required query expression. Individual queries are discussed in more detail in the Client Access Point (DAG-CAP) component descriptions for supported protocols.
Supported Query Types
The DAG system is designed to support fragment-matching queries on a limited set of data attributes -- "Name", "Organizational Role", "Organization", and "Locality". The selected permissible query combinations of attributes are listed in Table 3.1. From the table it can be seen that not all combinations of the three attributes are supported -- only those that are needed for the desired functionality.
Symbol Description ------- ----------- N Name NL Name + Locality NO Name + Organization NOL Name + Organization + Locality RO Role + Organization ROL Role + Organization + Locality
Table 3.1 DAG-supported queries
The RO and ROL queries are separated from the rest as they are searches for "virtual" persons -- roles within an organization (e.g., president, or customer service desk) for which one might want to find contact information.
Matching Semantics
As befits the individual client query protocols, more string matching expressions may be provided. The basic semantics of the DAG expect the following to be available in all client access software (as relevant):
Daigle & Hedberg Informational [Page 12]
RFC 2967 TISDAG October 2000
- Full word, exact match - Word substring match (E.g., "cat" would match "scatter") - Case-sensitive and case-insensitive matching
TISDAG: LDAP/X.500, supports case-sensitivity as such but some of the most used attributes, such as the commonName attribute, are defined in the standard to be of the case-insensitive attributetypes. The impact on the DAG system is that even if the index collected from a LDAP/X.500 server might have upper and lower case letters in the tokens, they can not be handled as such since that would be inferring meaning in something which is natively regarded as meaningless. The conclusion of the above is that The Referral Index should be case-insensitive and case- sensitivity should be supported by the SAPs if the native access protocol supports it.
Character Sets
Wherever possible, the DAG System supports and promotes the use of Unicode Version 2.0 for character sets (see [21]) specifically the UTF-8 encoding (see Appendix A.2 of [21] or [20]) Accommodation is made, where necessary, to support the deployed base of existing software.
Specifically:
DAG/IP: All internal communications using the DAG/IP are carried out in UTF-8.
TISDAG: not just UTF-8, but UTF-8 based on composed UNICODE version 2 character encodings.
DAG-CAP input: Where specific access protocols permit selection of character sets, DAG-CAPs must support UTF-8. They may additionally support other anticipated character set encodings.
DAG-SAP communications with WDSPs: Where specific access protocols permit selection of character sets, DAG-SAPs must support UTF-8 and use UTF-8 whenever the remote WDSP supports it. They may additionally support other character set encodings.
CIP Index Objects: The Index Objects supplied by the WDSPs to the DAG system shall contain data encoded in UTF-8.
TISDAG: The same limitation as for DAG/IP, that is the basic data should be UTF-8 encoded composed UNICODE version 2 character encodings.
The schema used for the DAG service is defined in Appendix A. This is a very basic information schema, intended to carry the necessary information for the DAG service, and not more. Although generic "whitepages" schema definitions do exist the more sophisticated and detailed the information presentation, the more difficult it is to map the schema seamlessly across protocols of different paradigms. Thus, the "KISS" ("Keep it simple, sir") principle seems appropriate here.
Individual DAG-CAPs define how they express this schema.
Referral Definition
For client access protocols that make use of the concept of referrals, DAG-CAP definitions will define the expression of referrals in those protocols. The DAG/IP defines the expression of referrals (see Appendix C).
Error conditions
Each DAG-CAP may provide more detailed error messages, but will define minimally the support for the following error conditions:
- unrecognized query - too many hits
Apart from these errors, the DAG-CAP may choose to refuse a query by redirecting the end-user to a different DAG-CAP of the same protocol.
The DAG will use the Common Indexing Protocol (CIP) server-server protocol to obtain updated index objects from WDSPs. For query- routing purposes, WDSPs are expected to provide Whois++, LDAPv2 or LDAPv3 interface to their data (although their preferred access may be something completely different). N.B.: In the responses from the technical survey, all respondents currently provide access to their service in one of these protocols.
In order to provide a useful and uniform service, WDSPs are expected to provide 7x24 access to their whitepages information. WDSPs are also expected to implement operations, administration, maintenance, and provisioning processes designed to minimize service down time for both planned and unplanned administration and maintenance activities.
The conceptual architecture of the DAG is represented in Figure 4.1. General architectural specifications are described below, followed by individual component specifications Sections 5.5 through 5.12.
Communications between components of the DAG will be by TCP/IP connections, using the DAG-Internal Protocol (DAG/IP). DAG/IP is used by DAG-CAPs to communicate with the Referral Index and DAG-SAPs. Thus, the DAG/IP defines
- the DAG-CAPs' range of query ability in the Referral Index (to gather referrals in response to the end-user's requests) - the responses (and their formats) of the Referral Index to the DAG-CAP requests - the DAG-CAPs' range of query ability to the DAG-SAPs for pursuing referrals when the DAG-CAP needs to do chaining for the client access software - the responses (and their formats) of the DAG-SAPs to the DAG-CAPs.
The detail of the planned DAG/IP is given in Appendix C. The detail of the DAG-CAP--Referral Index and DAG-CAP--DAG-SAP interactions is given in the definitions of individual DAG-CAPs and DAG-SAPs, below (Sections 5.5 through 5.12).
The Referral Index is responsible for maintaining the index of WDSP information, and providing a list of reasonable referrals in response to DAG-CAP search requests. These "referrals" provide pointers to identify WDSPs that may have information that matches the end-user's query.
Individual DAG-CAPs are responsible for providing a particular client access protocol interface to the DAG service. DAG-CAPs receive end- user queries in a particular query access protocol, convert the request into a query for the Referral Index ( i.e., expressed in DAG/IP), and then convert the Referral Index's response into a form that is appropriate for the client access protocol. This may mean passing back the referrals directly, calling on DAG-SAPs to do the work of translating the referral into results ("chaining"), or a combination of both.
Individual DAG-SAPs are called upon (by DAG-CAPs) to take DAG- generated referrals and pursue them -- issuing the indicated query at the specified WDSP service. Results from individual WDSPs are converted back into DAG/IP-specific format for the DAG-CAP that made the request. Each DAG-SAP is responsible for handling referrals to WDSPs of a particular protocol (e.g., LDAPv2, Whois++, etc).
This section notes some of the thinking that has driven the architectural and software design specification for the DAG system. This helps to provide the context in which to understand the software specifications that follow, and should give clues for the eventual extension of the DAG system. This section also acts, in some ways, as an FAQ (Frequently Asked Questions) section, as the content is shaped by questions received during the tech spec development phase. It attempts to illuminate context that may not otherwise be apparent on a first reading of the software specifications.
4.2.1 2 Distinct Functions: Referrals and Chaining
At all times, it must be kept in mind that the primary function of the DAG system is to provide users with referrals to WDSP services that may have the information they seek. Since it is the case that not all supported client protocols can handle referrals, the DAG system also provides a chaining service to pursue referrals that the user's client software cannot handle itself. This chaining service does attempt to match the user's query against data from WDSPs, but this is to be seen as a secondary, or support function of the DAG system. In the perfect future, all access protocols will be able to handle all referrals!
The DAG system does not attempt to be a chameleon, or the ultimate whitepages query service. It focuses on providing referrals for information on the limited number of query types outlined in the functional specifications of the DAG service. This makes the DAG system a good place to start a search, but refinements and detailed inquiries are beyond its scope.
Given the limited query syntax of the DAG system it will not always be possible to exactly match a query posed to a CAP into a query posed to a SAP. This will have the effect that for instance a LDAPv2
Daigle & Hedberg Informational [Page 17]
RFC 2967 TISDAG October 2000
client that issues a query to the DAG system which by the DAG system is chained to a LDAP server might not get the same results as if the client where directly connected to the server in question.
Even the limited query syntax of the DAG system is capable of expressing queries that might NOT be possible to represent in the access protocols to the WDSPs. In these cases the DAG-SAP either can refuse the query or try to emulate it.
As part of the chaining service offered by the DAG system, a certain amount of mapping between protocols is required -- in theoretical terms, there are "N" allowable end-user query access protocols, and "M" supported WDSP server protocols. The architecture of the software is constructed to use a single internal protocol (the DAG/IP) and data schema, providing a common language between all components. Without this, each input protocol module (DAG-CAP) would have to be constructed to be able to handle every WDSP protocol -- NxM protocol mappings. This would make the system complex, and difficult to expand to include new protocols in future.
4.2.6 DAG-CAPs and DAG-SAPs are completely independent of each other
For the above reasons, the DAG-CAP and DAG-SAP modules are intended to be completely independent of each other. A DAG-SAP responds to a query that is posed to it in the DAG/IP, without regard to the protocol of the DAG-CAP that passed the query.
Thus, the DAG-CAP is responsible for using the DAG/IP to obtain referral information and, where necessary, chained responses. Where necessary, it performs adjustments to accommodate the differences in semantics between the DAG/IP and its native protocol. This might involved doing post-filtering of the results returned by the DAG-SAPs since the query issued in DAG/IP to the DAG-SAP might be "broader" then the original query.
Thus, the DAG-CAP "knows" only 2 protocols: its native protocol, and the DAG/IP.
Similarly, the DAG-SAP is responsible for responding to DAG/IP queries by contacting the designated WDSP server. Where necessary, it performs adjustments to accommodate the differences in semantics between the DAG/IP and its native protocol. These adjustments might mean that, as a consequence, the DAG-SAP will receive results that do not match the original query. In such cases the DAG-SAP should attempt to do post-pruning in order to reduce the mismatch between the original query and the results returned.
Thus, the DAG-SAP "knows" only 2 protocols: its native protocol, and the DAG/IP.
No module outside of the DAG system should be aware of the DAG/IP's construction. End-users use the query protocols supported by DAG- CAPs; WDSPs are contacted using the query protocols supported in the DAG-SAPs.
The expectation is that the DAG system, although defined as a single construct, will operate by running modules on several different, perhaps widely distributed (in terms of geography and ownership), computers. For this reason, the DAG/IP specified in such a way that it will operate on inter-machine communications.
The DAG system architecture was constructed with a specific view to extensibility. At any time, an individual component may be improved (e.g., the Mail DAG-CAP may be given a different query interface) without disrupting the system.
Additionally, future versions of the DAG system may support other access protocols -- for end-users, and for WDSPs.
It is always a challenge to accurately represent text protocol in a printed document; when is a new line a "newline", and when is it an effect of the text formatter?
Daigle & Hedberg Informational [Page 19]
RFC 2967 TISDAG October 2000
In order to be adequately illustrated, this document includes many segments of protocol grammars, sample data, and sample input/output in a text protocol. In order to distinguish newlines that are significant in a protocol, the symbol
<NL>
is used. For example,
This is an example of a very long line of input. There is only one newline in it (at the end), in spite of the fact that this document shows it spanning several lines of text.<NL>
Every DAG-CAP must support the full range of DAG queries, as defined in 3.3.1.
Each DAG-CAP accepts queries in its native protocol. Individual DAG-CAP definitions define the expected expression of the DAG queries in the native protocol.
The DAG-CAP is then responsible for:
- converting that expression into a query in the DAG/IP to obtain relevant referrals from the Referral Index. This might mean that parts of the original query are disregarded (e.g., if the query included attributes not supported by the DAG application, or if the query algebra was not supported by the DAG application); - returning referrals in the client's native protocol, where possible; - expressing the client query to the necessary DAG-SAPs, given the limitations mentioned above, to chain those referrals not usefully expressible in the client's native protocol; - possibly doing post-filtering on the DAG-SAP results; and - converting the collected DAG-SAP results for expression in the client's native protocol (and schema, where applicable).
Each DAG-CAP defines the nature of the interaction with the end-user (e.g., synchronous or asynchronous, etc). Additionally, each DAG-CAP must be able to carry out the following, in order to permit load- limiting and load-balancing in the DAG system:
- direct the client to a different DAG-CAP of the same type (for load-balancing)
Daigle & Hedberg Informational [Page 20]
RFC 2967 TISDAG October 2000
- decline to return results because too many referrals were generated (to discourage data-mining). Ideally, this should include the generation of a message to refine the query in order to produce a more manageable number of referrals/replies.
DAG-CAPs must be capable of accepting and respecting DAG-SAP service referrals (for DAG-SAP load-sharing).
In protocols that permit it, the DAG-CAP should indicate to the end- user which services were unavailable for chaining referrals (i.e., to indicate there were parts of the search that could not be completed, and information might be missing).
TISDAG: Any CAP that receives commands other than queries, like help, answers those on its own. A CAP should not pass any system command on to the RI.
It must be possible to change the expected address of the DAG-CAP by configuration of the software (i.e., host and port, e-mail address, etc).
For DAG-CAPs that need to access DAG-SAPs for query chaining, for each type (protocol) of DAG-SAP that is needed, the DAG-CAP must be configurable in terms of:
- at least one known DAG-SAP of every necessary protocol to contact - for each DAG-SAP, the host and port of the DAG-SAP software
The DAG-CAPs must also be configurable in terms of a maximum number of referrals to handle for a user transaction (i.e., to prevent data mining, the DAG-CAP will refuse to reply if the query is too general and too many hits are generated at the Referral Index).
The DAG-CAP must be configurable in terms of alternate DAG-CAPs of the same type to which the end-user software may be directed if this one is too busy.
Apart from error conditions arising from the operation of the DAG-CAP itself, DAG-CAPs are responsible for communicating error conditions occurring elsewhere in the system that affect the outcome of the user's query (e.g., in the DAG-RI, or in one or more DAG-SAPs).
Daigle & Hedberg Informational [Page 21]
RFC 2967 TISDAG October 2000
If the DAG-CAP sends a query to the DAG-RI and receives an error message, it should attempt to match the the received DAG errorcode into its native access protocol's error codes. The same action is appropriate when the DAG-CAP is "chaining" the query to one DAG-SAP.
There are also occasions when the DAG-CAP may have to combine multiple errorcodes into a single expression to the user. When the DAG-CAP is "chaining" the query through DAG-SAPs to one or more WDSPs, situations can arise when there is a mix of responsecodes from the DAG-SAPs. If this happens, the DAG-CAP should try to forward information to the end-user software that is as specific as possible, for instance which of the WDSPs has not been able to fulfill the query and why.
See Appendix D for more information concerning error condition message mappings.
Since there is no perfect match between the query syntaxes of the DAG system on one hand and the different access protocols that the DAG- CAPs and DAG-SAPs supports on the other, there will be situations where the results a DAG-CAP has to collect is "broader" then what would have been the case if there had been a perfect match. This might have adverse effects on the system to the extent that administrative limits will "unnecessary" be exceeded on WDSPs or that the collected results exceeds the sizelimit of the DAG-CAP.
Since the DAG-CAP is the only part of the DAG system that actually knows what the original query was, the DAG-CAP can prune the results received from the DAG-SAPs in such a way that the results presented to the client better matches the original question.
Every DAG-SAP must support the full range of DAG queries, as defined in 3.3.1. Results must be complete DAG schemas expressed in well- formed DAG/IP result formats (see Appendix C). Each DAG-SAP accepts queries in DAG/IP and converts them to the native schema and protocol for which it is designed to proxy.
The DAG-SAP is then responsible for
- converting the query into the native schema and protocol of the WDSP to which the referral points. (If the query is not representable in the native protocol, it must return an error
Daigle & Hedberg Informational [Page 22]
RFC 2967 TISDAG October 2000
message. If it is emulatable, the DAG-SAP can attempt emulate it by posing a related query to the WDSP and post-pruning the results received); - contacting that WDSP, using the host, port, and protocol information provided in the referral; - negotiating the query with the remote WDSP; - accepting results from the WDSP, possibly doing post-filtering on the result set; and - conveying the results back to the calling DAG-CAP using the DAG/IP and its schema.
Note that this implicitly means that the DAG-SAP is responsible for chaining and pursuing any referrals it receives from WDSP services. The DAG-SAP returns only search results to the DAG-CAP that called it.
Since it might not be possible to exactly map a DAG query into a query in the access protocol supported by the a DAG-SAP, the DAG-SAP should try to translate it into a more general query (or if necessary into a set of queries). If so, the DAG-SAP must then prune the result set received before furthering it to the DAG-CAP.
Some constraints, search and case, can appear both as local and global constraints. If this happens in a query then the local constraint specification overrides the global. For a query like the following:
fn=leslie;search=exact and org=think:search=substring
the resulting search constraint for "fn=leslie" will be "exact" while it for "org=think" will be "substring".
The Referral Index contains (only) information necessary to deliver referrals to DAG-CAPs based on the query types supported by the DAG itself. The Referral Index creates an index over these objects so that it can respond to DAG-CAP queries using the DAG/IP. The information is drawn directly from interactions with participating WDSPs' software, using the Common Indexing Protocol (CIP).
WDSPs that wish to participate in the DAG system must register themselves (see Section 5.4.6). Once registered, the Referral Index will interact with the WDSPs using the Common Indexing Protocol as defined in [1], using the Index Object defined in Section 5.4.3.
The CIP index object type is based on the Tagged Index Object as defined in [12]. Appendix E details the expected content of the index objects as they are to be provided by the WDSPs.
TISDAG: The tokens in the Tagged Index Object should be UTF-8 encoded composed UNICODE version 2 character encoding.
The Referral Index interacts with the rest of the DAG internal modules (DAG-CAPs) by listening for queries and responding in the DAG/IP (defined in Appendix C).
The Referral Index must index the necessary attributes of the CIP index object in order to respond to queries of the form described in Table 3.1.
The semantics of the chosen CIP object (defined in Appendix E) are such that a referral to a WDSP server is sent back if (and only if)
- the index object of the WDSP contains all the tokens of the query, in the attributes specified, according to the logic of the DAG/IP query, and - all of those tokens are found with a common tag.
Daigle & Hedberg Informational [Page 24]
RFC 2967 TISDAG October 2000
This means that a query for the name "Fred Flintstone" (2 tokens) will yield a referral to a server that has a record for "Fred Amadeus Flintstone", but not to a WDSP with 2 differently tagged records, for "Fred Amadeus" and "Julie Flintstone". Depending on the access protocol being used and the original end-user query, the referral to the WDSP with "Fred Amadeus Flintstone" may yield a successful result, or it may not. But, it is known that the other WDSP would not have yielded successful searches. That is, the referral approach may yield false-positive results, but will not miss appropriate WDSPs.
The Referral Index will accept queries only from recognized (registered) DAG-CAPs. This will reduce "denial of service" attack types, but is also a reflection on the fact that the Referral Index uses the DAG/IP, (i.e., internal) protocol, which should not be exposed to non-DAG software.
The Referral Index must be able to use authenticated communication to receive data from WDSPs (see Appendix E).
This is the default Mail DAG-CAP. More sophisticated ones could certainly be written -- e.g., for pretty-printed output, or for handling different philosophies of case-matching.
This DAG-CAP has been designed on the assumption that mail queries will be human-generated (i.e., using a mail program/text editor), as opposed to being queries formulated by software agents. The input grammar should therefore be simple and liberal in acceptance of variations of whitespace formatting.
Mail DAG-CAP input is expected to be a regular or MIME-encoded (see [9] and [10]) SMTP mail message, sent to an advertised mail address. The mail DAG-CAP parses the message and replies to it with a MIME- encoded message containing the results of the DAG search.
One query is accepted per e-mail message -- text after a single valid query has been read is simply ignored.
The body of the query message must follow the syntax defined below. Note that all input control terms ("type=", "name=" etc) are shown in lower case for convenience, but could be upper case or mixed case on input.
A key element of translating from the Mail DAG-CAP input into the DAG/IP query format is to "tokenize" the input terms into single token elements for the DAG/IP query. For example, the n-term
name= thinking cat<NL>
is tokenized into 2 n-tokens:
thinking cat
which are then mapped into the following in the DAG/IP query (dag-n- terms):
FN=thinking and FN=cat<NL>
The same is true for all r-terms, l-terms and o-terms. The primary steps in translating the mail input into a DAG/IP query are:
translate quoted-printable encoding, if necessary translate base64 encoding, if necessary tokenize the strings for each term construct the DAG/IP query from the resulting components, as described in more detail below
DAG/IP constraints are constructed from the searchtype information in the query.
dag-matchtype = "search=" <matchtype> / "search=substring" ; if matchtype not ; specified
dag-casetype = "case=ignore" / ; if casetype not ; specified or ; casetype=ignore "case=consider" ; if casetype=sensitive
constraints = ":" dag-matchtype ";" dag-casetype
The terms for the DAG/IP query are constructed from the tokenized strings from the mail input.
Daigle & Hedberg Informational [Page 28]
RFC 2967 TISDAG October 2000
dag-n-terms = "FN=" n-token 0*( " and FN=" n-token) dag-o-terms = "ORG=" o-token 0*( " and ORG=" o-token) dag-l-terms = "LOC=" l-token 0*( " and LOC=" l-token) dag-r-terms = "ROLE=" r-token 0*( " and ROLE=" r-token)
This means that the relevant DAG/IP queries are formulated as one of two types:
n-query = dag-n-terms nl-query = dag-n-terms " and " dag-l-terms no-query = dag-n-terms " and " dag-o-terms nol-query = dag-n-terms " and " dag-o-terms " and " dag-l-terms ro-query = dag-r-terms " and " dag-o-terms rol-query = dag-r-terms " and " dag-o-terms " and " dag-l-terms
The examples given earlier are then translated as follows.
Example 1:
FN=thinking and FN=cat:search=substring;case=ignore<NL>
Example 2:
FN=thinking and FN=cat:search=exact;case=ignore<NL>
Example 3:
ROLE=thinking and ROLE=cat and ORG=space and ORG=colonization:search=substring;case=ignore<NL>
Querying a DAG-SAP
In querying a DAG-SAP (irrespective of the protocol of that DAG-SAP), the DAG/IP query must include information about the target WDSP server. This information is drawn from the Referral Index SERVER- TO-ASK referral information, and is appended to the query as specified in Appendix C):
Daigle & Hedberg Informational [Page 29]
RFC 2967 TISDAG October 2000
":host=" quoted-hostname ";port=" number ";server-info=" quoted-serverinfo ";charset=" charset
where the response from the Referral Index included:
and the "quoted-hostname" and "quoted-serverinfo" are obtained from "hostname" and "serverinfo" respectively, by quoting the DAG/IP special characters.
(N.B.: See Appendix C for further definitions of the terms used in the SERVER-TO-ASK response).
Note that it is the DAG-SAP's responsibility to extract these terms from the query and use them to identify the WDSP server to be contacted. See the individual DAG-SAP definitions, below.
The results message is sent to the "Reply-To:" address of the originating mail, if available (see [4] for appropriate interpretation of mail originator headers). The original query is repeated, along with the message-id. The remainder of the body of the mail message is the concatenation of responses from the DAG-SAP calls, each result having the WDSP's SOURCE URI (from the referral) appended to it, and the system messages also having been removed.
At the end of the message, the WDSP servers that failed to respond (i.e., the DAG-SAP handling the referral returned the "% 403 Information Unavailable" message) are listed with their server-info.
If the mail DAG-CAP receives a message that is not parsable using the query grammar described above, it returns an explanatory message to the query mail's reply address saying that the query could not be interpreted, and giving a description of valid queries.
If the number of referrals sent by the Referral Index is greater than the pre-determined maximum (for detecting data-mining efforts, or otherwise refusing over-general queries, such as "FN=svensson"), the mail DAG-CAP will send an explanatory message to the query mail's reply address describing the "over-generalized query" problem, suggesting the user resubmit a more precise query, and describing the list of valid query types.
If the mail DAG-CAP receives several different result codes from the DAG-SAPs it should represent those in an appropriate manner in the response message.
A mail DAG-CAP may redirect a connection to another mail DAG-CAP for reasons of load-balancing. This is done simply by forwarding the mail query to the address of the alternate mail DAG-CAP.
The web DAG-CAP provides its interface via standard HTTP protocol. The general expectation is that the web DAG-CAP will provide a form page with radio buttons to select "substring or exact match" and "consider case or ignore case". Other information (about name, role, organization, locality) is solicited as free-form text.
The DAG-CAP receives queries via an HTTP "post" method (the outcome of the form action for the page described above, or generated elsewhere). The rest of this section describes the variables that are to be expressed in that post. The actual layout of the page and most user interface issues are left to the discretion of the builder. Note that the Web DAG-CAP may be called upon to provide responses in different content encoding, and must therefore address the "Accept- Encoding:" request header in the HTTP connection.
Although the Web protocol, HTTP, is not itself capable of handling referrals, through the use of two extra variables this client is given the option of requesting referral information and then pursuing individual referrals through the Web DAG-CAP itself, as a proxy for those referrals. This is handled through the extra "control variables" to request referrals only, and to indicate when the transaction is a continuation of a previous query to pursue a referral.
There has been call to have a "machine-readable" version of the search output. As HTML is geared towards visual layout, user agents that intend to do something with the results other than present them in an HTML browser have few cues to use to extract the relevant information from the HTML page. Also, "minor" visual changes, accomplished with extensive HTML updates, can disrupt user agents that were built to blindly parse the original HTML. Therefore, provision has been made to return "raw" format results. These are requested by specifying "Accept-Content: application/whoispp- response" in the request header of the HTTP message to the HTTP DAG-CAP.
Daigle & Hedberg Informational [Page 32]
RFC 2967 TISDAG October 2000
The variables that are expected are:
transaction = "new" / "chain" ; default is "new". This ; should not be user-settable. It is used ; in constructed URLs resulttype = "all" / "referrals" ; default is "all" matchtype = "substring" / "exact" casetype = "case ignore" / "case sensitive" n-term = string o-term = string l-term = string r-term = string host-term = string port-term = string servinfo-term = string prot-term = string ; the protocol of the referral string = <UNICODE-2-0-UTF-8> / <UNICODE-1-1-UTF-8> / <ISO-8859-1>
If the transaction variable is "chain", the information in the POST is used to pursue a particular referral, not do a search of the Referral Index. The appropriate DAG-SAP (deduced from the prot-term) is contacted and issued the query directly.
Results from this type of query are always full results (i.e., not referrals).
Querying the Referral Index
A key element of translating from the Web DAG-CAP input into the DAG/IP query format is to "tokenize" the input terms into single token elements for the DAG/IP query. For example, the n-term
name= thinking cat
is tokenized into 2 n-tokens:
thinking cat
which are then mapped into the following in the DAG/IP query (dag-n- terms):
FN=thinking and FN=cat
Daigle & Hedberg Informational [Page 33]
RFC 2967 TISDAG October 2000
The same is true for the r-term, l-term and o-term.
The primary steps in translating the HTTP input into a DAG/IP query are:
translate encodings, if necessary tokenize the strings for each term construct the DAG/IP query from the resulting components, as described in more detail below
DAG/IP constraints are constructed from the searchtype information in the query.
dag-matchtype = "search=" <matchtype> / "search=substring" ; if matchtype not ; specified
dag-casetype = "case=ignore" / ; if casetype not ; specified or ; casetype="case ignore" "case=consider" ; if casetype= ; "case sensitive"
constraints = ":" dag-matchtype ";" dag-casetype
The terms for the DAG/IP query are constructed from the tokenized strings from the HTTP post input.
dag-n-terms = "FN=" n-token 0*( " and FN=" n-token) dag-o-terms = "ORG=" o-token 0*( " and ORG=" o-token) dag-l-terms = "LOC=" l-token 0*( " and LOC=" l-token) dag-r-terms = "ROLE=" r-token 0*( " and ROLE=" r-token)
This means that the relevant DAG/IP queries are formulated as one of two types:
nl-query = dag-n-terms " and " dag-l-terms no-query = dag-n-terms " and " dag-o-terms nol-query = dag-n-terms " and " dag-o-terms " and " dag-l-terms ro-query = dag-r-terms " and " dag-o-terms rol-query = dag-r-terms " and " dag-o-terms " and " dag-l-terms
Querying a DAG-SAP
In querying a DAG-SAP (irrespective of the protocol of that DAG-SAP), the DAG/IP query must include information about the target WDSP server. This information is drawn from the Referral Index SERVER- TO-ASK referral information, and is appended to the query as specified in Appendix C:
":host=" quoted-hostname ";port=" number ";server-info=" quoted-serverinfo ";charset=" charset
where the response from the Referral Index included:
and the "quoted-hostname" and "quoted-serverinfo" are obtained from "hostname" and "serverinfo" respectively, by quoting the DAG/IP special characters.
(N.B.: See Appendix C for further definitions of the terms used in the SERVER-TO-ASK response).
Note that it is the DAG-SAP's responsibility to extract these terms from the query and use them to identify the WDSP server to be contacted. See the individual DAG-SAP definitions, below.
If the resulttype was "all", all of the referrals received from the Referral Index are chained using the appropriate DAG-SAPs. If only referrals were requested, the Referral Index results are returned.
The default response encoding is text/html. If the resulttype was "all", the content of the chaining responses from the DAG-SAPs, without the system messages, is collated into a single page response, one result entry per demarcated line ( e.g., bullet item). The FN or ROLE value should be presented first and clearly. The SOURCE URI for each WDSP referral should be presented as an HREF for each of the WDSPs results.
At the end of the message, the WDSP servers that failed to respond (i.e., the DAG-SAP handling the referral returned the "% 403 Information Unavailable" message) are listed with their server-info.
If, however, the resulttype was "referrals", the results from the Referral Index are returned as HREF URLs to the Web DAG-CAP itself, with the necessary information to carry out the query (including the "HOST=", etc, for the referral).
For example, if the original query:
n-term="thinking cat" resulttype="referrals"
drew the following referral from the Referral Index:
The Source-URI should be established in the response as its own HREF URI.
application/whoispp-response Results
If Accept-Encoding: " HTTP request header had the value "application/whoispp-response", the content of the HTTP response will be constructed in the same syntax and attribute mapping as for the Whois++ DAG-CAP.
If the resulttype was "all", all the referrals will have been chained by the Web DAG-CAP, and the response will include only full data records.
If the resulttype was "referrals", then all referrals are passed directly back in a single response, in correct Whois++ referral format (conveniently, this is how they are formulated in the DAG/IP). Note that this will include referrals to LDAP-based services as well as Whois++ servers.
A Web DAG-CAP may redirect a connection to another web DAG-CAP for reasons of load-balancing. This is done simply by using an HTTP redirect.
Daigle & Hedberg Informational [Page 37]
RFC 2967 TISDAG October 2000
Standard Errors
If the web DAG-CAP receives a message that is not parsable using the query grammar described above, it sends an explanatory HTML page saying that the query could not be interpreted, and giving a description of valid queries.
If the number of referrals sent by the Referral Index is greater than the pre-determined maximum (for detecting data-mining efforts, or otherwise refusing over-general queries, such as "FN=svensson"), the web DAG-CAP will send a page with an explanatory message describing the "over-generalized query" problem, suggesting the user resubmit a more precise query, and describing the list of valid query types.
If the web DAG-CAP receives more than one result code from the DAG- SAPs, it must represent them all in a appropriate manner in the response.
application/whoispp-response Errors
An invalid query is responded to with a simple text response with the error: "% 500 Syntax Error".
If too many referrals are generated from the Referral Index, the simple text response will have the message "% 503 Query too general".
TISDAG: The system commands polled-for/-by should elicit the empty set as a return value until we better understand the implications of doing otherwise.
The Whois++ DAG-CAP formulates a DAG/IP query by forwarding the search terms received (as defined in Table 5.1).
For example, the above query would be expressed as:
FN=foo and LOC=kista and FN=bar and template=DAGPERSON<NL>
Querying a DAG-SAP
In querying a DAG-SAP (irrespective of the protocol of that DAG-SAP), the DAG/IP query must include information about the target WDSP server. This information is drawn from the Referral Index SERVER- TO-ASK referral information, and is appended to the query as specified in appendix C:
":host=" quoted-hostname ";port=" number ";server-info=" quoted-serverinfo ";charset=" charset
Daigle & Hedberg Informational [Page 39]
RFC 2967 TISDAG October 2000
where the response from the Referral Index included:
and the "quoted-hostname" and "quoted-serverinfo" are obtained from "hostname" and "serverinfo" respectively, by quoting the DAG/IP special characters.
(N.B.: See Appendix C for further definitions of the terms used in the SERVER-TO-ASK response).
Note that it is the DAG-SAP's responsibility to extract these terms from the query and use them to identify the WDSP server to be contacted. See the individual DAG-SAP definitions, below.
Results are expressed in Whois++ by collating the DAG/IP results received from DAG-SAPs (using the FULL response), and using the template and attribute mappings defined in Appendix B. For each result from a given referral, the SOURCE attribute is added, with the value of the SOURCE-URI from the referral.
Any referrals to other Whois++ servers provided by the Referral Index are sent directly to the Whois++ client as follows:
As appropriate, the Whois++ DAG-CAP will express operational errors following the Whois++ standard. There are 4 particular error conditions of the DAG system that the DAG-CAP will handle as described below.
When the Whois++ DAG-CAP receives a query that it cannot reply to within the (data) constraints of the DAG, it sends an error message and closes the connection. The error message includes
% 502 Search expression too complicated<NL>
If the number of referrals sent by the Referral Index is greater than the pre-determined maximum (for detecting data-mining efforts, or otherwise refusing over-general queries, such as "FN=svensson"), the Whois++ DAG-CAP will send an error message and close the connection. The error message includes
% 503 Query too general<NL>
(N.B.: this is different from the "Too many hits" reply, which does send partial results.)
Daigle & Hedberg Informational [Page 41]
RFC 2967 TISDAG October 2000
A Whois++ DAG-CAP may redirect a connection to another Whois++ DAG- CAP for reasons of load-balancing. This is expressed to the end-user client software using the SERVER-TO-ASK response with appropriate information to reach the designated alternate DAG-CAP.
If a Whois++ DAG-CAP receives several different response codes from DAG-SAPs it should try to represent them all in the response to the end-user client.
The proposed mapping between DAG/IP response codes and Whois++ response codes are given in Appendix D.
Input to the LDAPv2 DAG-CAP follows the LDAPv2 standard ([19]). Minimally, the LDAPv2 DAG-CAP must support the following queries (adapted from the ASN.1 grammar of the standard):
Filter ::= CHOICE { and [0] SET OF Filter, or [1] SET OF Filter, not [2] Filter, equalityMatch [3] AttributeValueAssertion, substrings [4] SubstringFilter }
SubstringFilter ::= SEQUENCE { type AttributeType, SEQUENCE OF CHOICE { initial [0] LDAPString, any [1] LDAPString, final [2] LDAPString } }
Queries against attributes in the prescribed LDAP standard schema (see Appendix B) are accepted.
N.B., this is a minimal set of supported queries, to achieve the basic DAG-defined queries. An LDAP DAG-CAP may choose to support more complex queries than this, if it undertakes to do the translation from the DAG/IP to the LDAPv2 client in a way that responds to the semantics of those queries.
TISDAG: Since LDAPv2 didn't specify any characterset but relied on X.500 to do so, in practice several different charactersets are in use in Sweden today. That the LDAPv2 CAP has no way of knowing which characterset that are in use by a connecting client is a problem that the TISDAG project can not solve.
Users of the DAG system will have to configure their specific client according to information on the TISDAG web page. That page provides very specific information (including port number) that can be given to LDAPv2 users. The LDAP DAG-CAP listening on the default port (389) will be the LDAPv3 one.
The essential stratagem for mapping LDAP queries into DAG/IP Referral Index queries is to tokenize the string-oriented LDAP AttributeValueAssertions or SubstringFilters and construct an appropriate DAG/IP token-oriented query in the DAG/IP. This will generalize the LDAP query and yield false-positive referrals, but should not miss any appropriate referrals.
There are 3 particular cases to be considered:
equalityMatch queries substring queries combination equalityMatch and substring queries
TISDAG: If the LDAP filter contains a cn-term and no objectclass specification it is unclear if the search is for a person or a role. When this happens the DAG query should cover all bases and map the query into a query for both people and roles.
EqualityMatch queries can be handled by simply tokenizing the AttributeValueAssertions, making one DAG/IP query term per token (using the appropriate DAGSchema attribute) and carrying out an exact match in the DAG/IP.
Consider the following example, represented in the ASCII expression of LDAP Filters as described in [13]):
(& (cn=Foo Bar)(objectclass=inetOrgPerson))
This query can be represented in the DAG/IP as
FN="Foo" and FN="Bar":search=exact<NL>
N.B. The search is set up to be "case=ignore" (the DAG/IP's default) because the relevant LDAP schema attributes are all derivatives of the "name" attribute element, which is defined to have a case insensitive match.
If no objectclass were defined the query in DAG/IP would have been
(FN="Foo" and FN="bar") or (ROLE="Foo" and ROLE="bar"):search=exact
Daigle & Hedberg Informational [Page 44]
RFC 2967 TISDAG October 2000
inetOrgPerson is used as the objectclass in this and the following examples, although person or organizationalPerson could also have been used.
This query will yield false-positive referrals; the original LDAP query should only match against records for which the "cn" attribute is exactly the phrase "Foo Bar", whereas the DAG/IP query will yield referrals any WDSP containing records that include the two tokens "foo" and "bar" in any order.
For example, this DAG/IP query will yield referrals to WDSPs with records including:
cn: Bar Foo cn: Le Bar Foo cn: Foo Bar AB
LDAP substring queries must also be tokenized in order to construct a DAG/IP query. The additional point to bear in mind is that LDAP substring expressions are directed at phrases, which obscure potential token boundaries. Consequently, all points between substring components must be considered as potential token boundaries.
could be expressed as a DAG/IP query with 3 tokens, in a substring search:
FN=black and ORG=c and ORG=t:search=substring<NL>
This query will yield false-positive results as the tokenized query does not preserve the order of appearance in the LDAP substring, and it doesn't preserve phrase-boundaries. That is,
ORG=c and ORG=t:search=substring
will match
tabacco
which is not a match by the LDAP query semantics.
Combined EqualityMatch and Substring queries need special attention. When an LDAP query includes both EqualityMatch components and substring filter components, the DAG/IP query to the Referral Index
Daigle & Hedberg Informational [Page 45]
RFC 2967 TISDAG October 2000
can be constructed by following the same mechanisms of tokenization, but the whole search will become a substring search, as the DAG/IP defines only search types across the entire query for Referral Index queries.
FN=Foo and FN=Bar and ORG=c and ORG=t:search=substring<NL>
Alternatively, the LDAP DAG-CAP could conduct two separate queries and take the intersection (the logical "AND") of the two sets of referrals returned by the Referral Index.
Note that DAG/IP can accept phrases for searches -- the query
FN=Foo\ bar<NL> (note the escaped space)
is perfectly valid. However, it would match only those things which have been tokenized in a way that preserves the space, which is the empty set in the case of the data stored here.
Querying a DAG-SAP
It is never invalid to use the same substantive query to a DAG-SAP as was used to obtain referral information from the Referral Index. However, the over-generalization of these queries may yield excessive numbers of results, and will necessitate some pruning of results in order to match the returned results against the semantics of the original LDAP query. It is the LDAP DAG-CAP that is responsible for this pruning, as it is the recipient of the original query, and responsible for responding to its semantics.
In concrete terms, when making the DAG/IP query which is to be sent to a DAG-SAP the above mentioned queries are still valid queries, but an alternative finer-grained query is also possible, namely:
FN=foo and FN=bar and ORG=c;search=lstring and ORG=t;search=tstring
Particularly in the case of the LDAPv2 DAG-CAP, however, there will be cause to use LDAP(v2/v3) DAG-SAPs. Since these DAG-SAPs also deal in phrase-oriented data, a less-over-generalized query can be passed to them:
FN=Foo\ Bar:search=exact<NL>
Daigle & Hedberg Informational [Page 46]
RFC 2967 TISDAG October 2000
In querying a DAG-SAP (irrespective of the protocol of that DAG-SAP), the DAG/IP query must include information about the target WDSP server. This information is drawn from the Referral Index SERVER- TO-ASK referral information, and is appended to the query as specified in Appendix C:
":host=" quoted-hostname ";port=" number ";server-info=" quoted-serverinfo ";charset=" charset
where the response from the Referral Index included:
and the "quoted-hostname" and "quoted-serverinfo" are obtained from "hostname" and "serverinfo" respectively, by quoting the DAG/IP special characters.
(N.B.: See Appendix C for further definitions of the terms used in the SERVER-TO-ASK response).
Note that it is the DAG-SAP's responsibility to extract these terms from the query and use them to identify the WDSP server to be contacted. See the individual DAG-SAP definitions, below.
As described above, results from DAG-SAPs will have to be post- processed in cases where the original query was generalized for expression in DAG/IP.
Acceptable results are expressed in the LDAP search response:
SearchResponse ::= CHOICE { entry [APPLICATION 4] SEQUENCE { objectName LDAPDN, attributes SEQUENCE OF SEQUENCE { AttributeType, SET OF AttributeValue } }, resultCode [APPLICATION 5] LDAPResult }
where
LDAPDN = DN / "cn=" (FN/ROLE) [",o="ORG] ",dc=se" attributes = <all attributes mapped from DAG schema, and "objectClass = inetOrgPerson", "objectClass = top", "objectClass = person" or "objectClass = organizationalRole", as appropriate, and "labeledURI = <SOURCE-URI>" for each result from a given referral>
(Where DN,FN,ORG and ROLE are the values from the DAG schema).
I.e., where available, the entry's true DN is used; otherwise (e.g., for data coming from Whois++ servers), a reasonable facsimile is constructed.
As appropriate, the LDAPv2 DAG-CAP will express system responses following the LDAPv2 standard.
Daigle & Hedberg Informational [Page 48]
RFC 2967 TISDAG October 2000
Appendix D gives the proposed mapping between DAG/IP response codes and LDAPv2 resultcodes.
There are 4 particular error conditions of the DAG system that the DAG-CAP will handle as described below.
When the LDAPv2 DAG-CAP receives a query that it cannot reply to within the (data) constraints of the DAG queries, it sends an error message and closes the connection. The error message includes the LDAPv2 resultCode:
noSuchAttribute (for incorrect schema attributes) inappropriateMatching (when a match type other than those supported is used, e.g. approxMatch) unwillingToPerform (when the query is not one of the defined types)
If the number of referrals sent by the Referral Index is greater than the pre-determined maximum (for detecting data-mining efforts, or otherwise refusing over-general queries, such as "FN=svensson"), the LDAPv2 DAG-CAP will send an error message. The error message includes one of the following resultCodes:
sizeLimitExceeded timeLimitExceeded
An LDAPv2 DAG-CAP may redirect a connection to another LDAPv2 DAG-CAP for reasons of load-balancing. This is expressed to the end-user client software using the "umich referral" convention to direct the client software to an alternate DAG-CAP by passing the URL in an error message.
Since a LDAPv2 DAG-CAP only can send one resultcode back to a client; If a LDAPv2 DAG-CAP receives several different result codes from the DAG-SAPs it will have to construct a resultmessage that to some extent represents the combination of those. It is proposed that in these cases the following actions are taken:
- All the response codes are collected - Each response code are translated into the corresponding LDAPv2 resultcode. - A resultcode is chosen to represent the collected response on the following grounds: If "success" is the only resultcode represented after these steps the return that result code. If apart from "success" there is one other resultcode represented return that other resultcode.
Daigle & Hedberg Informational [Page 49]
RFC 2967 TISDAG October 2000
If apart from "success" there are two or more resultcodes represented return the resultcode "other".
Input to the LDAPv3 DAG-CAP follows the LDAPv3 definition (currently defined in [17]). Minimally, the LDAPv3 DAG-CAP must support the following queries (adapted from the ASN.1 grammar of the standard):
BindRequest ::= [APPLICATION 0] SEQUENCE {
version INTEGER (1 .. 127), name LDAPDN, authentication AuthenticationChoice }
Filter ::= CHOICE { and [0] SET OF Filter, or [1] SET OF Filter, not [2] Filter, equalityMatch [3] AttributeValueAssertion, substrings [4] SubstringFilter }
SubstringFilter ::= SEQUENCE { type AttributeDescription, -- at least one must be present substrings initial [0] LDAPString, substrings any [1] LDAPString, substrings final [2] LDAPString}
Queries against attributes in the proscribed LDAP standard schema (see Appendix B) are accepted.
N.B., this is a minimal set of supported queries, to achieve the basic DAG-defined queries. An LDAP DAG-CAP may choose to support more complex queries than this, if it undertakes to do the translation from the DAG/IP to the LDAPv3 client in a way that responds to the semantics of those queries.
The essential stratagem for mapping LDAP queries into DAG/IP Referral Index queries is to tokenize the string-oriented LDAP AttributeValueAssertions or SubstringFilters and construct an appropriate DAG/IP token-oriented query in the DAGschema. This will generalize the LDAP query and yield false-positive referrals, but should not miss any appropriate referrals.
There are 3 particular cases to be considered:
equalityMatch queries substring queries combination equalityMatch and substring queries
TISDAG: If the LDAP filter contains a cn-term and no objectclass specification it is unclear if the search is for a person or a role. When this happens the DAG query should cover all bases and map the query into a query for both people and roles.
Daigle & Hedberg Informational [Page 51]
RFC 2967 TISDAG October 2000
EqualityMatch queries can be handled by simply tokenizing the AttributeValueAssertions, making one DAG/IP query term per token (using the appropriate DAGSchema attribute) and carrying out an exact match in the DAG/IP.
Consider the following example, represented in the ASCII expression of LDAP Filters as described in [13]):
(& (cn=Foo Bar)(objectclass=person))
This query can be represented in the DAG/IP as
FN="Foo" and FN="Bar":search=exact<NL>
N.B. The search is set up to be "case=ignore" (the DAG/IP's default) because the relevant LDAP schema attributes are all derivatives of the "name" attribute element, which is defined to have a case insensitive match.
If no objectclass where defined the query in DAG/IP would have been
(FN="Foo" and FN="bar") or ( ROLE="Foo" and ROLE="bar"):search=exact
Although person is used as objectclass in this and the following examples, inetOrgPerson or organizationalPerson could also have been used.
This query will yield false-positive referrals; the original LDAP query should only match against records for which the "cn" attribute is exactly the phrase "Foo Bar", whereas the DAG/IP query will yield referrals any WDSP containing records that include the two tokens "foo" and "bar" in any order.
For example, this DAG/IP query will yield referrals to WDSPs with records including:
cn: Bar Foo cn: Le Bar Foo cn: Foo Bar AB
LDAP substring queries must also be tokenized in order to construct a DAG/IP query. The additional point to bear in mind is that LDAP substring expressions are directed at phrases, which obscure potential token boundaries. Consequently, all points between substring components must be considered as potential token boundaries.
Daigle & Hedberg Informational [Page 52]
RFC 2967 TISDAG October 2000
Thus, the LDAP query
(& (cn=black) o=c*t) (objectclass=person))
should be expressed as a DAG/IP query with 3 tokens, in a substring search:
FN=black and ORG=c and ORG=t:search=substring<NL>
This query will yield false-positive results as the tokenized query does not preserve the order of appearance in the LDAP substring, and it doesn't preserve phrase-boundaries. That is,
ORG=c and ORG=t:search=substring
will match
tabacco
which is not a match by the LDAP query semantics.
Combined EqualityMatch and Substring queries need special attention. When an LDAP query includes both EqualityMatch components and substring filter components, the DAG/IP query to the Referral Index can be constructed by following the same mechanisms of tokenization, but the whole search will become a substring search, as the DAG/IP defines search types across the entire query.
Thus,
(& (cn=Foo Bar) (o=c*t) (objectclass=person))
can be expressed as
FN=Foo and FN=Bar and ORG=c and ORG=t:search=substring<NL>
Alternatively, the LDAP DAG-CAP could conduct two separate queries and take the intersection (the logical "AND") of the two sets of referrals returned by the Referral Index.
Note that DAG/IP can accept phrases for searches -- the query
FN=Foo\ bar<NL> (note the escaped space)
is perfectly valid. However, it would match only those things which have been tokenized in a way that preserves the space, which is the empty set in the case of the data stored here.
Daigle & Hedberg Informational [Page 53]
RFC 2967 TISDAG October 2000
Querying a DAG-SAP
It is never invalid to use the same substantive query to a DAG-SAP as was used to obtain referral information from the Referral Index. However, the over-generalization of these queries may yield excessive numbers of results, and will necessitate some pruning of results in order to match the returned results against the semantics of the original LDAP query. It is the LDAP DAG-CAP that is responsible for this pruning, as it is the recipient of the original query, and responsible for responding to its semantics.
In concrete terms, when making the DAG/IP query which is to be sent to a DAG-SAP the above mentioned queries are still valid queries, but an alternative finer-grained query is also possible, namely:
FN=foo and FN=bar and ORG=c;search=lstring and ORG=t;search=tstring
In querying a DAG-SAP (irrespective of the protocol of that DAG-SAP), the DAG/IP query must include information about the target WDSP server. This information is drawn from the Referral Index SERVER- TO-ASK referral information, and is appended to the query as specified in Appendix C):
"host=" quoted-hostname ";port=" number ";server-info=" quoted-serverinfo ";charset=" charset
where the response from the Referral Index included: "# SERVER-TO-ASK " serverhandle <NL> " Server-info: " serverinfo<NL> " Host-Name: " hostname<NL> " Host-Port: " number<NL> " Protocol: " prot<NL> " Source-URI: " source<NL> " Charset: " charset<NL> "# END"<NL>
and the "quoted-hostname" and "quoted-serverinfo" are obtained from "hostname" and "serverinfo" respectively, by quoting the DAG/IP special characters.
(N.B.: See Appendix C for further definitions of the terms used in the SERVER-TO-ASK response).
Note that it is the DAG-SAP's responsibility to extract these terms from the query and use them to identify the WDSP server to be contacted. See the individual DAG-SAP definitions, below.
As described above, results from DAG-SAPs will have to be post- processed in cases where the original query was generalized for expression in DAG/IP. Acceptable results are expressed in LDAPv3 messages containing search result entries (see the standard for more detail):
attributes = <all attributes mapped from the DAG schema, and "objectClass = inetOrgPerson", "objectClass = person", "objectClass = top" or "objectClass = organizationalRole", as appropriate, and "labeledURI = <SOURCE-URI>" for each result from a given referral> LDAPResult = success
(Where DN, FN, ROLE, and ORG are the values from the DAG schema).
I.e., where available, the entry's true DN is used; otherwise (e.g., for data coming from Whois++ servers), a reasonable facsimile is constructed.
Referral URLs are constructed from the DAG/IP's SERVER-TO-ASK information as follows:
The intention is that WDSPs using LDAPv3 servers will provide an appropriate LDAPDN for their server in the SERVER-INFO. Clients are then expected to repeat their query at the server designated by this URL (i.e., the refURL does not include the query).
As appropriate, the LDAPv3 DAG-CAP will express operational errors following the LDAPv3 standard. There are 4 particular error conditions of the DAG system that the DAG-CAP will handle as described below.
When the LDAPv3 DAG-CAP receives a query that it cannot reply to within the (data) constraints of the DAG queries, it sends an error message and closes the connection. The error message includes the LDAPv3 resultCode
noSuchAttribute (for incorrect schema attributes chosen) inappropriateMatching (when a match type other than those supported is used e.g., approxMatch) unwillingToPerform (when the query is not one of the defined types)
If the number of referrals sent by the Referral Index is greater than the pre-determined maximum (for detecting data-mining efforts, or otherwise refusing over-general queries, such as "FN=svensson"), the LDAPv3 DAG-CAP will send an error message. The error message includes the following resultCode:
Daigle & Hedberg Informational [Page 56]
RFC 2967 TISDAG October 2000
adminLimitExceeded
An LDAPv3 DAG-CAP may redirect a connection to another LDAPv3 DAG-CAP for reasons of load-balancing. In this case, the LDAPv3 DAG-CAP sends a result message including only
SearchResultReference ::= [APPLICATION 19] AltURL
SearchResultDone ::= referral
where
AltURL = "ldap://" <althostport> ":" <altbase>
Since a LDAPv3 DAG-CAP only can send one resultcode back to a client; If a LDAPv3 DAG-CAP receives several different result codes from the DAG-SAPs it will have to construct a resultmessage that to some extent represents the combination of those. It is proposed that in these cases the following actions are taken:
- All the response codes are collected - Each response code are translated into the corresponding LDAPv3 resultcode. - A resultcode is chosen to represent the collected response on the following grounds: If "success" is the only resultcode represented after these steps the return that result code. If apart from "success" there is one other resultcode represented return that other resultcode. If apart from "success" there are two or more resultcodes represented return the resultcode "other".
The Whois++ DAG-SAP expects valid DAG/IP communications. Queries must include referral information (see below) and search terms that conform to the DAG-allowed query types (e.g., not searches for organization alone, etc).
The referral information is added to the end of the DAG-SAP query, as defined in the DAG-CAP definition sections:
":host=" quoted-hostname ";port=" number ";server-info=" quoted-serverinfo ";charset=" charset
The HOST and PORT information are used to make a TCP/IP-based connection to the remote (presumed) Whois++ server. The query expressed to the remote Whois++ server is the remainder of the DAG/IP query the Whois++ DAG-SAP received, with the following template ID translations:
template=DAGPERSON becomes template=USER
and
template=DAGROLE becomes template=ORGROLE
Additional mappings for attributes are defined in Appendix B.
Note that the search types used in the DAG/IP are not all required by the Whois++ syntax. Therefore, some Whois++ WDSPs may be using servers that do not support searches other than "exact" and "lstring" (the search types required by the Whois++ protocol standard). The Whois++ DAG-CAP may
- send the DAG/IP query as constructed (e.g., with "search=substring"), and pass back the "% 502 Search expression too complicated" from the WDSP's server, - translate the DAG/IP query into a construct using only these search types (which will yield incomplete results, as not all queries are expressible with those search types), - attempt to ascertain what search types are supported by the remote server and reformulate using them (e.g., regular expressions). This would work, but would entail an excessively complicated Whois++ DAG-SAP, and might not yield any better results if the remote server doesn't support any optional search types.
Any referrals that the remote WDSP server returns are pursued, following the usual Whois++ (client) fashion, by the Whois++ DAG-SAP.
If it is not possible to establish a Whois++ session with the remote server, or if the session is interrupted, before results are received, the DAG-SAP will itself return no results and an error message, including
% 403 Information Unavailable<NL>
Daigle & Hedberg Informational [Page 58]
RFC 2967 TISDAG October 2000
If the remote server issues any other Whois++ error message and does not yield any results, the remote server's error message will be included in the DAG-SAP's own error message; no results will be returned.
If results are successfully received from the remote server, they will be expressed using the DAG/IP -- essentially passing through all FULL response information received from the remote server, mapped into the DAGSchema using the mappings defined in Appendix A.
The LDAPv2 DAG-SAP expects valid DAG/IP communications. Queries must include referral information (see below) and search terms that conform to the DAG-allowed query types (e.g., not searches for organization alone, etc).
The referral information is added to the end of the DAG-SAP query, as defined in the DAG-CAP definition sections (as additional terms in the DAG/IP query):
":host=" quoted-hostname ";port=" number ";server-info=" quoted-serverinfo ";charset=" charset
The HOST and PORT information are used to make a TCP/IP-based connection to the remote (presumed) LDAPv2 server. The DAG-SAP will establish a connection with the remote server, following standard LDAPv2 message exchanges.
The search request itself will be constructed from the DAG/IP query (without the HOST, SERVER-INFO and PORT terms) as follows:
timeLimit INTEGER (0 .. maxInt), attrsOnly FALSE filter Filter, attributes SEQUENCE OF AttributeType -- all DAGschema attributes equivalents in the defined standard LDAP schema }
Filter ::= CHOICE { and [0] SET OF Filter, or [1] SET OF Filter, not [2] Filter, substrings [4] SubstringFilter, }
SubstringFilter SEQUENCE { type AttributeType,
SEQUENCE OF CHOICE { substrings initial [0] LDAPString, substrings any [1] LDAPString, substrings final [2] LDAPString} }
where and, or and not filters are constructed to preserve the logic of the DAG/IP query.
For the purposes of matching token-based DAG/IP queries to reasonable LDAP queries, all searches should be passed to the LDAP WDSP as substring searches. The WDSP results must then be pruned to respect token boundaries, where necessary.
So, for example, the DAG/IP query
FN=Foo\ Bar and ORG=Thinking\ Cat:search=substring<NL>
but the WDSPs returned results would have to be pruned to remove any results that had non-tokenizing characters on either side of "Foo Bar" and "Thinking Cat".
The final consideration for mapping DAG/IP queries into LDAP queries is the issue of character case. In LDAP, individual attribute syntaxes define the consideration of case. All of the attributes used here are case-insensitive in their definitions. Therefore, all LDAP WDSP queries are inherently case-insensitive; if the DAG/IP query calls for a case-sensitive match, the LDAP DAG-SAP will have to do pruning of the results from the DAG-SAP.
If it is not possible to establish an LDAPv2 session with the remote server, or if the session is interrupted before results are received, or if the remote server issues any kind of error message and produces no result, the DAG-SAP will itself return no results and an error message, including
% 403 Information Unavailable<NL>
If results are successfully received from the remote server, the attributes and values that are provided for each result message will be incorporated into the DAG/IP result, according to the schema mappings laid out in Appendix B.
One particular adjustment must be done to accommodate differences between LDAP and the DAG/IP. The attributes on which searches are keyed ("cn", "l", and "o" in the LDAP schemas) are all defined as being case-insensitive for equality matching. Thus, if the DAG/IP query includes the constraint "case=consider", the results from the remote server must be post-processed to remove any wrong-cased ones.
TISDAG: The serverhandle and localhandle in the DAG/IP response should be constructed as follows:
serverhandle is: <hostname-without-periods><port> (because server DN's are not enforceably unique). E.g., a services.bunyip.com server on 7778 would become servicesbunyipcom7778. localhandle is: the RDN (relative distinguished name), with spaces replaced by "_". E.g., cn=leslie_daigle
The LDAPv3 DAG-SAP expects valid DAG/IP communications. Queries must include referral information (see below) and search terms that conform to the DAG-allowed query types (e.g., not searches for organization alone, etc).
The referral information is added to the end of the DAG-SAP query, as defined in the DAG-CAP definition sections:
":host=" quoted-hostname ";port=" number ";server-info=" quoted-serverinfo ";charset=" charset
The HOST and PORT information are used to make a TCP/IP-based connection to the remote (presumed) LDAPv3 server. The DAG-SAP will establish a connection with the remote server, following standard LDAPv3 message exchanges.
The search request itself will be constructed from the DAG/IP query (without the HOST, SERVER-INFO and PORT terms) as follows:
SearchRequest ::= [APPLICATION 3] SEQUENCE { baseObject LDAPDN, -- from the DAG/IP query scope baseObject (0) }, derefAliases ENUMERATED { neverDerefAliases (0), derefInSearching (1), derefFindingBaseObj (2), derefAlways (3) }, sizeLimit INTEGER (0 .. maxInt), timeLimit INTEGER (0 .. maxInt), attrsOnly FALSE filter Filter, attributes SEQUENCE OF AttributeType -- all DAGschema attributes equivalents in the defined standard LDAP schema }
Filter ::= CHOICE { and [0] SET OF Filter, or [1] SET OF Filter,
Daigle & Hedberg Informational [Page 62]
RFC 2967 TISDAG October 2000
not [2] Filter, substrings [4] SubstringFilter, }
SubstringFilter SEQUENCE { type AttributeType, SEQUENCE OF CHOICE { substrings initial [0] LDAPString, substrings any [1] LDAPString, substrings final [2] LDAPString} }
where and, or and not filters are constructed to preserve the logic of the DAG/IP query.
For the purposes of matching token-based DAG/IP queries to reasonable LDAP queries, all searches should be passed to the LDAP WDSP as substring searches. The WDSP results must then be pruned to respect token boundaries, where necessary.
So, for example, the DAG/IP query
FN=Foo\ Bar and ORG=Thinking\ Cat:search=substring<NL>
but the WDSP's returned results would have to be pruned to remove any results that had non-tokenizing characters on either side of "Foo Bar" and "Thinking Cat".
The final consideration for mapping DAG/IP queries into LDAP queries is the issue of character case. In LDAP, individual attribute syntaxes define the consideration of case. All of the attributes used here are case-insensitive in their definitions. Therefore, all LDAP WDSP queries are inherently case-insensitive; if the DAG/IP query calls for a case-sensitive match, the LDAP DAG-SAP will have to do pruning of the results from the DAG-SAP.
Any referrals that the remote WDSP server returns are pursued, following the usual LDAPv3 (client) fashion, by the LDAPv3 DAG-SAP.
If it is not possible to establish an LDAPv3 session with the remote server, or if the session is interrupted before results are received, or if the remote server issues any kind of error message and produces no result, the DAG-SAP will itself return no results and an error message, including
% 403 Information Unavailable<NL>
If results are successfully received from the remote server, the attributes and values that are provided for each result message will be incorporated into the DAG/IP result, which will be expressed using the DAG/IP and schema mappings as outlined in Appendix A.
One particular adjustment must be done to accommodate differences between LDAP and the DAG/IP. The attributes on which searches are keyed ("cn", "l", and "o" in the LDAP schemas) are all defined as being case-insensitive for equality matching. Thus, if the DAG/IP query includes the constraint "case=consider", the results from the remote server must be post-processed to remove any wrong-cased ones.
TISDAG: The serverhandle and localhandle in the DAG/IP response should be constructed as follows:
- serverhandle is: <hostname-without-periods><port> (because server DN's are not enforceably unique). E.g., a services.bunyip.com server on 7778 would become servicesbunyipcom7778. - localhandle is: the RDN (relative distinguished name), with spaces replaced by "_". E.g., cn=leslie_daigle
The following sample end-user queries illustrate some of the more delicate steps of query/schema semantics translations in the DAG system.
N.B.: the data presented in these examples is often senseless, provided only to serve as illustrations of matching on word-ordering, case sensitivity, etc.
In this example, the Whois++ DAG-CAP receives the following query:
name=thinking and name=cat:search=exact;case=consider<NL>
The expected answer can be described as:
Any USER templates that contain the tokens "thinking" and "cat" in a name attribute.
For example:
Different records:
name: the thinking cat name: sublime cat thinking
or a single record with 2 or more name attributes
name: thinking felines name: erudite cat
but not
name: Thinking Cat Enterprises
This last record would not match because the query called for case sensitivity, and the case of the name attribute's value does not match the query.
What the Whois++ DAG-CAP sends to the Referral Index
After schema translation, this is sent to the Referral Index as:
fn=thinking and fn=cat:search=exact<NL>
What the Whois++ DAG-CAP Sends to an LDAP DAG-SAP
Note that the Whois++ DAG-CAP will never interact with a Whois++ DAG-SAP as the Whois++ referrals returned by the Referral Index are passed directly back to the Whois++ client.
The Whois++ DAG-CAP should send the same substantive query to the DAG-SAP as it sent to the Referral Index, except that it can include the case sensitivity constraint:
Daigle & Hedberg Informational [Page 65]
RFC 2967 TISDAG October 2000
fn=thinking and fn=cat:search=exact;case=consider<NL>
which will be translated by the DAG-SAP into an LDAP query of the form:
What the LDAP user is looking for, with this query, is all records within the "green groceries" organization that have a cn attribute starting with "th", ending with "t", and having a "c" somewhere in the middle.
cn values that would match this include:
cn: thinkingcat cn: Thinking Cat cn: The Black Cat cn: Thick Mat
Daigle & Hedberg Informational [Page 66]
RFC 2967 TISDAG October 2000
5.13.3 What the LDAP DAG-CAP sends to the Referral Index
The LDAP DAG-CAP must formulate a token-based query to the Referral Index that will not inadvertently exclude records that would match. The first challenge lies in the fact that the "*" characters in the LDAP string-based query can cover token-boundaries.
A suitable query to the Referral Index would be:
FN=th AND FN=C AND FN=T AND ORG=green AND ORG=groceries:search=substring<NL>
This will generate some false positive referrals, directing the query to WDSPs containing records with the following attribute values (the match letters are in capitals for ease of identification):
cn: wiTH three blaCk poTs
o: peaGREEN and cyan GROCERIES o: GROCERIES are GREENer than electronics
Alternative approaches include breaking the original query into several queries to the referral index in such a way that the DAG-CAP can use only those referrals that appear in all the Referral Index responses. However, this is
overkill -- the purpose of the Referral Index is to give direction on where there may be more information
difficult to code into the DAG-CAP in a general way -- it has to identify, by LDAP query type, when and how to do so
likely to generate Referral Index queries that are complex and time- consuming to process.
What the LDAP DAG-CAP Sends to a Whois++ DAG-SAP
The LDAP DAG-CAP may send the same query to a Whois++ DAG-SAP as it sent to the Referral Index. False positives here mean results that are not expected as a match by the LDAP client. The LDAP DAG-CAP should prune these results from the information returned by the Whois++ DAG-SAP.
Or it might rewrite the query into:
FN=th;search=lstring AND FN=C;search=substring AND FN=T;search=tstring AND ORG=green AND ORG=groceries:case=ignore<NL>
Daigle & Hedberg Informational [Page 67]
RFC 2967 TISDAG October 2000
What the LDAP DAG-CAP Sends to an LDAP DAG-SAP
As an architectural principle, it is never wrong to send the same query to a DAG-SAP as was formulated for the Referral Index. It is also noteworthy to keep in memory that all DAG-SAPs are handled equal by all DAG-CAPs therefore a LDAP DAG-CAP will not need to send a different query to a LDAP DAG-SAP then it would to any other DAG-SAP.
So in this case the LDAP DAP-CAP could either send the same query to the LDAP DAG-SAP as it sent to the Referral Index or it could send the augmented version that is allowed to be use with the DAG-SAPs, namely:
FN=th;search=lstring AND FN=C;search=substring AND FN=T;search=tstring AND ORG=green\ groceries:case=ignore<NL>
Note that this will be translated, by the LDAP DAG-SAP, into a query of the form
which is still more general than the original query.
Note the translation from "FN=th;search=lstring" into "cn=*th*". This is necessary, as the DAG/IP lstring constraint is based on tokens, whereas "cn=th*" refers to the beginning of the attribute's value (phrase, not token). The DAG-SAP should therefore prune out any results that include things like "oTHer plaCes for visiTors" in order to match the semantics of the DAG/IP query it received.
The DAG-CAP should then prune those results to match the semantics of the original LDAP query.
To satisfy the requirements laid out for the TISDAG project, the software built for the DAG system must be able to meet the following service specifications:
- primary designated DAG-CAPs of all types (but not necessarily secondary ones set up for load-balancing) must be available to provide service or redirect queries on a 7x24 basis.
Daigle & Hedberg Informational [Page 68]
RFC 2967 TISDAG October 2000
- in general, responses to queries should be available in under 10 seconds; very generalized queries (i.e., when the user truly cannot specify enough information to focus the search) can be deferred to take much longer (having results is more important than having a quick answer) - the data provided from each WDSP should be updated in the DAG at least once every 7 days
WDSPs who wish to participate in the DAG system do so by providing DAG-compatible access to their service, where DAG-compatible means:
- access in (exactly) one of LDAPv2, LDAPv3, or Whois++ - 7x24 service for responding to referrals generated in the DAG core (minimally) weekly updates of the index object describing the information their service indexes - use of USER and ROLE templates for Whois++ servers - use of inetorgperson and organizationalrole objectclasses for LDAP servers
To participate, WDSPs must register each DAG-compliant server with the DAG system, providing details for each data set that it covers:
- the host, port and protocol of the server - an identifier for the dataset - a URL for the service of preference for accessing the data (preferred source) - protocol-specific information - administrative contact information - CIP object exchange information
Note that any WDSP wishing to make data available through the DAG system but unable to support these requirements may provide information through an agreement with a third-party which does meet these requirements. Thus, data can be replicated between cooperating WDSPs. The DAG referral index does not claim ownership of personal information; it directs queries to services that do, by whatever agreements with whichever relevant parties. Note that, in this case, the SOURCE-URI may direct end-users to the WDSP's existing services, not the service of the third party.
It is anticipated that the DAG system will be quite popular, and measures must be available to distribute the load of answering queries.
Daigle & Hedberg Informational [Page 69]
RFC 2967 TISDAG October 2000
The DAG system is presented as a conceptual whole, made up of several component parts -- DAG-CAPs, DAG-SAPs and the Referral Index. Each of these component parts must be replicable, and service must be shared between replicas.
It may be interesting to consider allowing large-scale service providers (large companies, ISPs) the ability to mirror the Referral Index or provide alternate DAG-CAPs/DAG-SAPs for their personnel/customers. Policies and possibilities for doing that are beyond the scope of this report; however, the software architecture has been designed to support such activity.
Figure 6.1 shows that individual components of the DAG system may each run on non-co-located server hardware, connected by TCP/IP networks. These components can be replicated as needed.
Thus, the software built to this specification must be configurable to permit the following actions:
- DAG-CAP software must be able to handle or redistribute the primary load. Depending on the DAG-CAP software, this may be handled by having multiple processes attending to incoming queries, or the DAG-CAP at the primary address for the protocol may be nothing more than a reflector that redirects incoming queries to the address of the least-loaded server at the moment. - This is particularly necessary in synchronous connection protocols, such as Whois++ and LDAP, where the goal is to minimize the amount of time a requesting client is connected to the well-advertised address port. - DAG-CAP software must be able to direct referrals to different DAG-SAPs of the same protocol type. - DAG-CAP software must be able to detect overly general queries (i.e., have some metric to decide that the number of referrals generated by the Referral Index is too great). - DAG-SAPs must be able to redirect DAG-CAP queries at their discretion, or just refuse service because of loading (therefore DAG-CAPs must also be able to find other DAG-SAPs)
The DAG system has been designed to allow for extensibility in certain key areas:
It is possible to add new DAG-CAPs and DAG-SAPs transparently. Beyond replicating the software of existing DAG-CAPs, new implementations for particular protocols (e.g., building a more elaborate mail-based query system), or implementations for altogether different protocols (e.g., PH) can be added by adhering to the basic principles of DAG-CAPs and DAG-SAPs defined in the software specification. The new DAG-CAP is responsible for the translation of queries into DAG/IP (post-processing results, if necessary) and results in the new protocol. No other part of the DAG system is affected.
More functionality may be added to the DAG system service (e.g., adding security certificate references to the schema of returned information) by updating the DAG schema.
Depending on how the load on the service goes, it may be interesting to consider reducing the number of queries that are chained for protocols that inherently can handle the concept of pursuing referrals. Specifically, LDAPv3 and Whois++ both handle referrals, but the current system calls for chaining LDAPv3 (and LDAPv2) referrals for the Whois++ DAG-CAP, and vice versa. Alternatively,
Daigle & Hedberg Informational [Page 72]
RFC 2967 TISDAG October 2000
"virtual" DAG-CAPs could be established for each participating WDSP for each protocol the WDSP doesn't support, and referrals to those DAG-CAPs could be given to the calling client. For example, a Whois++ client would be given a Whois++ referral to the virtual Whois++ DAG-CAP for a WDSP that supports only LDAP. The importance of having one virtual DAG-CAP per WDSP is that the point of connection is the only way to distinguish which WDSP the Whois++ client thought it was connecting to.
Security, in the context of "read-only" directory services, is primarily concerned with maintaining data integrity as it passes from an originating server to the end-user making an inquiry. That is, some server(s) hold correct user information, and a client accessing a directory service should be certain that whichever servers that the information has to pass through before reaching the client, it receives a true representation of the original information.
The DAG system as such MUST be completely invisible as the mediator of the information from the WDSPs to the querying directory access client. The only possible modifications that can appear is translations from one characterset into another. Hopefully, this does not alter the meaning of the information.
In keeping with the public nature of the proposed TISDAG service, the DAG system does not provide any access control system beyond components' configuration to accept connections from recognized other components. For more detailed access control, it is up to the connected WDSPs to apply the access control.
Since the DAG system only supports searching and retrieving information, no updates can occur through the DAG client access points.
Security in updates (CIP index objects) is provided by encryption and signature of objects from registered WDSPs.
This work came from ideas originally put forward by Patrik Faltstrom. The TISDAG project was supported by the Swedish KK Foundation.
Thanks to especially to Jens Lundstrom, Thommy Eklof, Bjorn Larsson and Sandro Mazzucato for their comments on draft versions of this document.
Daigle & Hedberg Informational [Page 74]
RFC 2967 TISDAG October 2000
Appendix A - DAG Schema Definitions
The DAG makes use of 2 information schemas -- the DAGPERSON schema for information about specific people, and the DAGORGROLE schema for organizational roles that may or may not be job positions occupied by people at any given time (e.g., an organization's president, customer service desk, etc).
This appendix defines the schemas in terms of the attributes used within the DAG/IP. Mappings to the standard LDAP and Whois++ object classes and templates (respectively) are described in Appendix B.
Because the role of the DAG schemas is to act as an intermediary between information provided in different access protocols, with different underlying schema paradigms, the attributes in the schema are identified as being required or optional. The required attributes are so designated because they are involved in the DAG search types and/or the minimal returned response. They have defined mappings in the selected access protocols. The optional attributes have proposed mappings in those protocols.
It is important to note that the DAG/IP is constructed to carry any alternative attribute information that may be provided by a given WDSP; individual DAG-SAPs and DAG-CAPs may choose to pass along, interpret, or ignore any attributes not defined in this appendix.
Additionally, note that the order of attributes in the DAG/IP is significant, which means that it is possible to use one attribute to carry the information describing the type of subsequent ones (e.g., see the "ADR-TYPE" attribute below).
Finally, attributes may be repeated. For example, this schema structure can carry multiple phone numbers of different types for one person.
Daigle & Hedberg Informational [Page 75]
RFC 2967 TISDAG October 2000
A.1 DAG Personal Information Schema (DAGPERSON Schema)
Attribute Designation Specific Description --------- ----------- ------------------------------------- FN Required Free-text representation of full name EMAIL Required Internet e-mail address LOC Required Locality -- geographic region ORG Required Person's organization ADR-TYPE Optional Type of address that follows ("org", "home", "org-postal", "home-postal", "unqualified") ADR Optional Full address ADR-STREET Optional Street address component ADR-ROOM Optional Suite or room number component ADR-CITY Optional City name ADR-STATE Optional Region of address ADR-COUNTRY Optional Country ADR-CODE Optional Postal code component TEL-TYPE Optional Type of telephone number ( "work", "home", "mobile", "fax" ,"pager", "unqualified") in the following attribute TEL Optional A phone number for the person SOURCE Optional The WDSP's preferred access to their service -- a URL DN Optional Entry's "distinguished name" (for LDAP)
Table A.1 DAGPERSON schema attributes
Daigle & Hedberg Informational [Page 76]
RFC 2967 TISDAG October 2000
A.2 DAG Organizational Role Information Schema (DAGORGROLE Schema)
Attribute Designation Specific Description --------- ----------- --------------------- ROLE Required Name of organizational role EMAIL Required E-mail address associated with role ORG Required Name of organization LOC Required Locality -- geographic region TEL-TYPE Optional Type of telephone number in the TEL attribute immediately following("org" or "fax") TEL Optional Phone number
FN Optional Full name of current role occupant SOURCE Optional The WDSP's preferred access to their service -- a URL DN Optional Entry's "distinguished name" (for LDAP)
Table A.2 DAGORGROLE schema attributes
Appendix B - Schema Mappings for Whois++ and LDAP
The DAG/IP makes use of two specific schemas, as defined above. However, schemas particular to access protocols need to be handled in order to appropriately address incoming user queries, and chaining queries to WDSPs. The recognized standard schemas are:
- the USER template for Whois++ ([8]) - the ORGROLE template for Whois++ ([8]) - the inetOrgperson objectclass for LDAP ([16]) - the organizationalrole objectclass for LDAP ([18])
The DAG/IP schemas were developed based on the information that the TISDAG project requirements wish to return in results, in conjunction with information about standard schemas used in the basic WDSP access protocols (LDAPv2/v3 and Whois++). However, particularly in the case of address information, the schemas used for those protocols allow for considerable scope of information representation. In practice, this means that different WDSPs may choose to use different sub-parts of the schema, or even implement local customizations.
Therefore, Appendix A outlines a very basic schema that can carry all the necessary information. The basic DAG-CAPs and DAG-SAPs are designed to work to that information structure. This appendix outlines the expected behaviour for DAG-SAPs mapping into the DAG/IP schema, and DAG-CAPs extracting information to pass along to client software after a chaining operation has returned results.
The only time information is carried in the DAG schemas is when a DAG-SAP is returning information (obtained from WDSPs' servers) to a DAG-CAP using the DAG/IP. The "canonical" mappings between standard LDAP object classes (inetorgPerson, defined in [16] and organizationalRole, defined in [18] and the DAGPERSON schema and DAGORGROLE schema are defined such that information passed from an LDAP DAG-SAP to an LDAP DAG-CAP (e.g., in the case of an LDAPv3 DAG- SAP returning information chained for an LDAPv2 DAG-CAP) will be mapped into the same attributes as it was extracted.
However, the representation of some attributes (such as address) is truly widely varied between protocol paradigms. The goal with the "reasonable approximation" mappings that are provided is to give DAG-CAPs a basic mechanism for communicating information drawn from non-LDAP DAG-SAP sources. The mappings may not be perfect, but they will convey the information to the end-user in some LDAP- understandable fashion, which is the goal of this project's effort.
The canonical mappings for the LDAP inetorgPerson object class and the DAGPERSON schema are given in Table B.1. A few reasonable approximation mappings follow in Table B.2. Beyond that, DAG-SAPs may pass along any additional attributes in the DAG/IP, and DAG-CAPs may elect to forward or interpret any that are recognizable (e.g., the sn ("surname") attribute is not listed here, but a DAG-SAP might return that in the DAG/IP, and a DAG-CAP, recognizing the string representation, could elect to include it in its LDAP response to the client).
DAGPERSON Attribute LDAP inetorgPerson attribute ------------------- ---------------------------- FN cn EMAIL mail LOC l ORG o
ADR-TYPE=org ADR-STREET street ADR-ROOM roomNumber ADR-STATE st ADR-COUNTRY c
DAGROLE Attribute LDAP organizationalRole attribute ----------------------- --------------------------------- ADR-TYPE=unqualified ADR street ADR-STREET street ADR-ROOM room ADR-STATE st ADR-COUNTRY c
TEL-TYPE=unqualified TEL telephoneNumber
Table B.2 Reasonable Approximations for LDAP organizationalRole attributes
Daigle & Hedberg Informational [Page 79]
RFC 2967 TISDAG October 2000
For example, consider the following LDAP record information, in LDIF [11] format:
dn: cn=Barbara Jensen, ou=Product Development, o=Ace Industry, c=US objectclass: top objectclass: person objectclass: organizationalPerson objectclass: inetorgperson cn: Barbara Jensen cn: Barbara J Jensen cn: Babs Jensen sn: Jensen uid: bjensen telephonenumber: +1 408 5551212 description: A big sailing fan
This would validly be carried in the DAGPERSON schema as follows:
DN: cn=Barbara Jensen, ou=Product Development, o=Ace Industry, c=US FN: Barbara Jensen FN: Barbara J Jensen FN: Babs Jensen SN: Jensen TEL-TYPE: work TEL: +1 408 5551212
The canonical mappings for the LDAP organizationalRole object class and the DAGORGROLE schema are given in Table B.3 .Beyond that, DAG- SAPs may elect to send along any attributes, and DAG-CAPs may interpret any that are recognizable. N.B., the organizationalRole class does not include provision for inclusion of an e-mail address. This mapping rather blithely assumes the availability of the mail attribute as defined for inetorgPerson.
Daigle & Hedberg Informational [Page 80]
RFC 2967 TISDAG October 2000
DAGORGROLE Attribute LDAP organizationalRole attribute -------------------- --------------------------------- ROLE cn EMAIL mail ORG o LOC l
TEL-TYPE=org TEL telephoneNumber
TEL-TYPE=fax TEL facsimileNumber
FN roleOccupant DN dn SOURCE labeledURI
Table B.3 Canonical mappings for LDAP organizationalRole attributes
The "canonical" mappings between standard Whois++ templates as defined in [8] and the DAGPERSON schema and DAGORGROLE schema are defined in Tables B.4 and B.5. Beyond that, DAG-SAPs may pass along any additional attributes in the DAG/IP, and DAG-CAPs may elect to forward or interpret any that are recognizable.
DAGPERSON Attribute Whois++ USER template attribute ------------------- ------------------------------- FN name EMAIL email LOC address-locality ORG organization-name
The use of the DAG/IP is strictly internal to the DAG system. In that regard, it is possible make use of any query language, or define a new one.
The Whois++ protocol was selected as the basis of the DAG/IP for several reasons:
- it has the power and flexibility to convey all necessary queries - it is a simple, text-based protocol; clients need not implement the full functionality of the protocol in order to carry out minimal queries - the power of the full-fledge directory service query protocol will give DAG-CAP writers the ability to express more sophisticated queries if desired (e.g., to produce more intricate "intelligent" matching of spellings, common character substitutions, etc). - the text-based, delimited attribute results expression facilitates optional inclusion of extra data supplied by WDSPs -- DAG-CAPs can easily ignore any unknown information and continue to interpret the rest of the result information.
Also, the use of an existing protocol leverages the experience and time of the creators of the protocol -- hammering out such elusive and yet necessary details as handling line-endings, quoting special characters, etc.
There is a freely-available test suite of tools for testing servers' Whois++ protocol conformance (for the Referral Index, and for DAG- SAPs). Send mail to digger-info@bunyip.com for further information.
Input interactions in DAG/IP are as defined in RFC1835, "Architecture of the WHOIS++ service" ([6]), sections 2.2 and 2.3. Section C.3 of this document adapts the grammar used in more recent descriptions of the Whois++ protocol to illustrate the syntax of the DAG/IP.
DAG/IP output will be a subset of what is defined in RFC1835, section 2.4, except that referral responses ("SERVER-TO-ASK") contain more information.
The following sections are adapted from the Whois++ grammar. For discussion of the semantic intent of the query protocol, and other matters, see Whois++ RFC 1835 [6].
The following grammar, which uses the Augmented BNF (ABNF) notation as defined in [5], defines the set of acceptable DAG/IP input.
N.B.: As outlined in the ABNF definition, rule names and string literals are in the US-ASCII character set, and are case-insensitive. Also, when a character is written explicitly in the grammar, as for example ";", it represents the byte value of that character in all of the allowed character sets in their encodings used in this protocol. Specifically in UNICODE, ";" means the character U+003B, which when encoding the character in UTF-8 will generate the byte value 0x3B which is then used in the DAG/IP protocol.
TISDAG: Since the DAG system only supports certain attribute combinations in its queries, (Table 3.1). The use of generalterm may lead to unexpected behaviour and is therefore deprecated. CAPs should therefore not use it even if it is in the protocol.
; N.B.: If an attribute is named both with the "include" and "ignore" ; constraints, the attribute is to be included in the result, but the ; system message must be "% 112 Requested constraint not fulfilled".
TISDAG: Within the boundaries of the TISDAG project it has been decided that the only permitted attributes for wdspinfo are "host","port","server-info" and "charset". Regarding "charset" the values for this attribute are defined to be one of "UTF-8", "ISO8859-1","T\.61" or "US-ASCII".
datastring = 1*data-elt
attributename = 1*(<%d32-126 except specialbyte>) ; omit 127, which is DEL
NOTE: Spaces (sp) that are significant to a query must be escaped. The following characters, when significant to the query, may be preceded and/or followed by a single space: : ; , ( ) = !
The following grammar, which uses the Augmented BNF (ABNF) notation as defined in RFC2234 (see [5]),
N.B.: As outlined in the ABNF definition, rule names and string literals are in the US-ASCII character set, and are case-insensitive. Also, when a character is written explicitely in the grammar, as for example ";", it represents the byte value of that character in all of the allowed character sets in their encodings used in this protocol. Specifically in UNICODE, ";" means the character U+003B which when encoding the character in UTF-8 will generate the byte value 0x3B which is then used in the DAG/IP protocol.
- Whois++, whatever the server-handle on the record returned by the WDSP. - LDAP, <hostname-without-periods><port> (because server DN's are not enforceably unique). E.g., a services.bunyip.com server on 7778 would become servicesbunyipcom7778.
localhandle is: - Whois++: the localhandle on the record returned by the WDSP - LDAP, it is the RDN (relative distinguished name), with spaces replaced by "_". E.g., cn=leslie_daigle
The following list and discussion of response codes is derived from the Whois++ protocol definition, RFC1835 ([6]).
A system message begins with a '%', followed by a space and a three digit number, a space, and an optional text message. The line message must be no more than 81 bytes long, including the terminating CR LF pair. There is no limit to the number of system messages that may be generated.
A multiline system message have a hyphen instead of a space in column 6, immediately after the numeric response code in all lines, except the last one, where the space is used.
Example 1
% 200 Command okay
Example 2
Daigle & Hedberg Informational [Page 89]
RFC 2967 TISDAG October 2000
% 220-Welcome to % 220-the Whois++ server % 220 at ACME inc.
The client is not expected to parse the text part of the response message except when receiving reply 600 or 601, in which case the text part is in the former case the name of a character set that will be used by the server in the rest of the response, and in the latter case when it specifies what language the attribute value is in. The valid values for characters sets is specified in the "characterset" list in the BNF listing in Appendix C.
The theory of reply codes is described in Appendix E in STD 10, RFC821 ([15]).
System response code Description
---------------------------- ------------------------------ 110 Too many hits The number of matches exceeded the value specified by the maxhits constraint. Server will still reply with as many records as "maxhits" allows.
111 Requested constraint not One or more constraints in query supported is not implemented, but the search is still done.
112 Requested constraint not One or more constraints in query fulfilled has unacceptable value and was therefore not used, but the search is still done.
200 Command Ok Command accepted and executed. The client must wait for a transaction end system message.
201 Command Completed Command accepted and executed. successfully
502 Search expression too This message is sent when the complicated server is not able to resolve a query (i.e. when a client sent a regular expression that is too deeply nested).
503 Query to general This is like the "too many hits" situation, but the server does not send along any results. This message is used to deflect data mining.
600 <token> Subsequent attribute values are encoded in the character set specified by <token>.
601 <token> Subsequent attribute values are in the language specified by <token>.
Daigle & Hedberg Informational [Page 91]
RFC 2967 TISDAG October 2000
601 DEF Subsequent attribute values are default values, i.e. they should be used for all languages not specified by "601 <token>" since last "601 ANY" message.
601 ANY Subsequent attribute values are for all languages.
Table D.1 LDAPv2/v3 resultcodes to DAG/IP response codes mapping
Daigle & Hedberg Informational [Page 93]
RFC 2967 TISDAG October 2000
DAG/IP LDAP v2/v3 --------------------------------------- -------------------------- 110 Too many hits sizeLimitExceeded (4) 111 Requested constraint not supported constraintViolation (19) 112 Requested constraint not fullfilled constraintViolation (19) 200 Command Ok Success (0) 201 Command Completed successfully N.A. 203 Bye N.A. 204 Overgeneralized N.A. 220 Service Ready N.A. 226 Transaction complete N.A. 401 Service not available unavailable (52) 402 Search expression too complicated unwillingToPerform (53) 403 Information Unavailable busy (51) 404 Time out timeLimitExceeded (3) 405 Operations error operationsError (1) 500 Syntax error protocolError (2) 502 Search expression too complicated unwillingToPerform (53) 503 Query to general unwillingToPerform (53) 505 Operations error operationsError (1) 600 <token> N.A. 601 <token> N.A. 601 DEF N.A. 601 ANY N.A.
Table D.2 Mapping from DAG/IP response codes to LDAPv2/v3 resultcodes
DAG/IP Whois++ -------------------------------------- ----------------------------- 110 Too Many hits 110 Too Many hits 111 Requested constraint not supported 111 Requested constraint not supported 112 Requested constraint not fullfilled 112 Requested constraint not fullfilled 200 Command Ok 200 Command Ok 201 Command Completed successfully 201 Command Completed successfully 401 Service not available 401 Service not available 403 Information Unavailable 403 Information not available 404 Timeout 404 Timeout 405 Operations error 405 Operations error 500 Syntax error 500 Syntax error 502 Search expression too complicated 502 Search expression too complicated 503 Query to general 506 Query to general 505 Operations error 505 Operations error
Table D.3 Mapping between DAG/IP and Whois++ response codes
The CIP object used by the DAG system is based on the Tagged Index Object as defined in [12]. The grammar, adapted from that Work in Progress, for the specific object used by the DAG is as follows:
N.B.: The only tokenization type permitted is "TOKEN". While the Tagged Index Object memo permits the use of "FULL" (i.e., the entire value of the attribute is preserved as a single token), that has the danger of yielding a unique token for every record. Studies in the growth of centroid sizes as a function of number of records (see [14]) demonstrate that such unique tokens (e.g., phone numbers) are to be avoided. While storing tag information requires some number of extra bytes of storage per token index entry, using unique tokens causes the number of token entries in the index to continue to grow linearly with the number of records, thereby affecting search efficiency.
Note also that tags are to be applied to the data on a per entry level. Thus, if two index lines in the same index object contain the same tag, then it is always the case that those two lines refer back to the same "record" in the directory. In LDAP terminology, the two lines would refer back to the same directory object.
Daigle & Hedberg Informational [Page 96]
RFC 2967 TISDAG October 2000
Additionally if two index lines in the same index object contain different tags, then it is always the case that those two lines refer back to different records in the directory.
The attribute "objectclass" is used to denote the record/object types in the data summarized in this index object.
Values for the objectclass attribute should be restricted to: dagperson or dagrole, the two DAG schema object types.
WDSPs are expected to create index objects following the general principles outlined in the Whois++ protocol documentation (creation of centroids) and the Tagged Index Object documentation ([12]). Following the syntax described above, the index object contains token information for each attribute in the DAGSchema:
- a list of all the unique tokens (strings delimited by the specified characters) that appear in the WDSP database for the attribute - for each token in that list, which records the token appears in
So, for example,
Record #1: FN: Foo Bar ORG: The Snack Bar
Record #2: FN: Bar Smith ORG: Snack Shack
yields (conceptually) the following information for the attribute FN:
Foo (1), Bar (1,2), Smith (2)
and the following information for the attribute ORG:
The (1), Snack (1, 2), Bar (1), Shack (2)
Note that the record numbers here are used simply as tags or virtual record identifiers to indicate when 2 tokens appear in the same record. The record identifiers are not used for any part of any query to the WDSP.
Daigle & Hedberg Informational [Page 97]
RFC 2967 TISDAG October 2000
There is some discussion as to whether the use of the same record tag for all attributes makes it too easy to "decompile" the index object; i.e., reconstruct a WDSPs data based on re-ordering the tokens associated with each attribute and tag number. However, we are dealing only with the search attributes here, which is a minimal subset of the quantity of data held by the WDSP. The conclusion is then that the improved efficiency given by using the same tag numbers across attributes outweighs the (remote) possibility of information reconstruction.
This would yield the index object:
version: x-tagged-index-1 update-type: total this-update: 855938804
last-update: context-size: BEGIN IO-Schema objectclass: TOKEN
FN: TOKEN ORG: TOKEN END IO-Schema BEGIN Index-Info objectclass: */dagperson FN: 1/Foo -1,2/Bar -2/Smith ORG: 1/The -1,2/Snack -1/Bar -2/Shack End Index-Info
TISDAG: Within the project it has been decided to base consistency between updates on consistent tags. This means that if the update-type is "incremental" the specifier must be "tagbased".
It is beyond the scope of this document to define how WDSP servers shall be registered with the DAG Referral Index. Such a procedure must be defined, and the following information established for each WDSP dataset (adapted from the Tagged Index Object specification, [12]):
Daigle & Hedberg Informational [Page 98]
RFC 2967 TISDAG October 2000
dsi: An OID which uniquely identifies the subtree and scope of the dataset for which the index object is created.
base-uri: One or more URI's which will form the base of any referrals created based upon the index object that is governed by this agreement. For example, for LDAP the base-uri would specify (among other items): the LDAP host, the base object to which this index object refers (e.g., c=SE), and the scope of the index object (e.g., single container).
supplier: The hostname and listening port number of the supplier server, as well as any alternative servers holding that same naming contexts, in case the supplier is unavailable.
source-uri: The URI of the WDSP's preferred source of directory service information. This might be, for instance, an HTTP-based service.
consumeraddr: This is a URI of the "mailto:" form, with the RFC 822 email address of the consumer server.
updateinterval: The maximum duration in seconds between occurrences of the supplier server generating an update. If the consumer server has not received an update from the supplier server after waiting this long since the previous update, it is likely that the index information is now out of date. A typical value for a server with frequent updates would be 604800 seconds, or every week.
attributeNamespace: Every set of index servers that together wants to support a specific usage of indices, has to agree on which attributenames to use in the index objects. The participating directory servers also has to agree on the mapping from local attributenames to the attributenames used in the index. Since one specific index server might be involved in several such sets, it has to have some way to connect a update to the proper set of indexes. One possible solution to this would be to use different DSIs.
consistencybase: How consistency of the index is maintained over incremental updates: complete - every change or delete concerning one object has to contain all tokens connected to that object. This method must be supported by any server who wants to comply with this standard tagbased - starting at a full update every incremental update referring back to this full updated has to maintain state- information regarding tags, such that a object within the original database is assigned the same tagnumber every time. This method is optional.
Daigle & Hedberg Informational [Page 99]
RFC 2967 TISDAG October 2000
uniqueID - every object in the Dataset has to have a unique value for a specific attribute in the index. A example of such a attribute could be the distinguishedName attribute. This method is also optional.
securityoption: Whether and how the supplier server should sign and encrypt the update before sending it to the consumer server. Options for this version of the DAG service are "none": the update is sent in plaintext "PGP/MIME": the update is digitally signed and encrypted using PGP (see [7]). PGP/MIME is recommended.
security credentials: The long-term cryptographic credentials used for key exchange and authentication of the consumer and supplier servers, if a security option was selected. For "PGP/MIME", this will be the trusted public keys of both servers.
CIP Index Objects are sent to the DAG Referral Index by MIME-encoded SMTP, following the Common Indexing Protocol specification (see [2] and [3]).
Appendix F - Summary of Technical Survey Results
As part of the TISDAG project, a technical survey was carried out -- announced on the tisdag@swip.net mailing list, all Swedish WDSPs (and potential WDSPs) were encouraged to fill out and submit the WWW-based survey form (see http://tisdag.sunet.se/tisdag-survey.html).
The survey was carried out in May, 1997. Response was not as good as had been hoped -- in the end, 5 WDSPs participated. We had hoped for more responses than this, in order to have a concrete sense of directory service providers' current and planned status. However, informal "hallway" conversations with a few people at Interoperabilitet'97 in Sollentuna suggest that, while people see the TISDAG project as an important and timely step, they don't necessarily have an immediate understanding of how it will impact them, and what they can/should contribute. So, the results can be seen as informational, though not a definitive statement of the whole directory service picture in Sweden.
Interesting things to note from these results include the fact that, although there were only 5 respondents, these are clearly significant players -- 4 expect to have more than 100 000 records to contribute by 12 months from now. There were no real surprises in terms of the supported protocols or search types.
Daigle & Hedberg Informational [Page 100]
RFC 2967 TISDAG October 2000
Table E.1 summarizes information from the survey concerning types of queries currently supported by WDSPs, and planned for the next 12 months. Note that, at the time of the survey, the requirement of searching by ROLE had not been proposed, so the survey did not specifically ask if WDSPs supported both the DAGPERSON schema protocol-equivalents (i.e., USER template in Whois++ and inetorgperson objectclass in LDAP). In the table, the column "Complete info?" describes whether or not the WDSP currently returns at least as much information as is required for a DAG reply.
Table F.2 Summary of TISDAG Survey Results: Operational Information
Daigle & Hedberg Informational [Page 101]
RFC 2967 TISDAG October 2000
Appendix G - Useful References
N.B.: The following is a collection of Internet standards documents (RFCs) and Internet-Drafts from which the material in this report was drawn. Internet-Drafts are works-in-progress, and are not meant to be cited. Where they are used in this document, references are to the text contained in the Internet-Draft; i.e., they are not meant to imply standards, so much as useful starting points for the work of this project.
Electronic copies of the version of the Internet-Drafts documents that were used in preparing this report are available from the project web page, http://tisdag.sunet.se.
Bibliography
[1] Allen, J. and M. Mealling, "The Architecture of the Common Indexing Protocol", RFC 2651, August 1999.
[2] Allen, J. and M. Mealing, "MIME Object Definitions for the Common Indexing Protocol (CIP)", RFC 2652, August 1999.
[3] Allen, J. and P. Leach, "CIP Transport Protocols", RFC 2653, August 1999.
[4] Crocker, D., "Standard for the Format of ARPA Internet Text Messages", STD 11, RFC 822, August 1982.
[5] Crocker, D., "Augmented BNF for Syntax Specifications: ABNF", RFC 2234, November 1997.
[6] Deutsch, P., Schoultz, R., Falstrom, P. and C. Weider, "Architecture of the WHOIS++ Service", RFC 1835, July 1995.
[7] Elkins, M., "MIME Security with Pretty Good Privacy (PGP)", RFC 2015, October 1996.
[8] Patrik Faltstrom, Martin Hamilton, Leslie L. Daigle, "WHOIS++ templates", Work in Progress.
[9] Freed, N. and N. Borenstein, "Multipurpose Internet Mail Extensions (MIME) Part One: Format of Interent Message Bodies", RFC 2045, November 1996.
[10] Freed, N. and N. Borenstein, "Multipurpose Internet Mail Extensions (MIME) Part Two: Media Types", RFC 2046, November 1996.
Daigle & Hedberg Informational [Page 102]
RFC 2967 TISDAG October 2000
[11] Good, G., "The LDAP Data Interchange Format (LDIF) - Technical Specification", RFC 2849, June 2000.
[12] Hedberg, R., Greenblatt, B., Moats, R. and M. Wahl, "A Tagged Index Object for use in the Common Indexing Protocol", RFC 2654, August 1999.
[13] Howes, R., "A String Representation of LDAP Search Filters", RFC 1960, June 1996.
[14] Paul Panotzki, "Complexity of the Common Indexing Protocol: Predicting Search Times in Index Server Meshes", Master's Thesis, KTH, September 1996.
[15] Postel, J., "Simple Mail Transfer Protocol", STD 10, RFC 821, August 1982.
[16] Smith, M., "Definition of the inetOrgPerson Object Class", RFC 2798, April 2000.
[17] Wahl, M., Howes, T. and S. Kille, "Lightweight Directory Access Protocol (v3)", RFC 2251, December 1997.
[18] Wahl, M., "A summary of the X.500(96) User Schema for use with LDAPv3", RFC 2256, December 1997.
[19] Yeong, W., Howes, T. and S. Kille, "Lightweight Directory Access Protocol", RFC 1777, March 1995.
[20] Yergeau, F., "UTF-8, a transformation format of ISO 10646", RFC 2279, January 1998.
[21] The Unicode Consortium, "The Unicode Standard -- Version 2.0", Addison-Wesley, 1996.
Daigle & Hedberg Informational [Page 103]
RFC 2967 TISDAG October 2000
Authors' Addresses
Leslie L. Daigle Thinking Cat Enterprises
EMail: leslie@thinkingcat.com
Roland Hedberg Catalogix Jegerveien 25 0777 Oslo Norway
Phone: +47 23 08 29 96 EMail: Roland@catalogix.se
Daigle & Hedberg Informational [Page 104]
RFC 2967 TISDAG October 2000
Full Copyright Statement
Copyright (C) The Internet Society (2000). 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.