[Docs] [txt|pdf] [Tracker] [Email] [Nits]

Versions: 00

INTERNET-DRAFT                                                Y. Arrouye
August 1, 2001                                                 V. Parikh
Expires February 1, 2002                                         N. Popp
                                                         RealNames Corp.

          Keyword Lookup Systems As a Class of Naming Systems
                        draft-arrouye-kls-00.txt

Status of this Memo

   This document is an Internet-Draft and is in full conformance with
   all provisions of Section 10 of RFC2026.

   Internet-Drafts are working documents of the Internet Engineering
   Task Force (IETF), its areas, and its working groups. Note that other
   groups may also distribute working documents as Internet-Drafts.

   Internet-Drafts are draft documents valid for a maximum of six months
   and may be updated, replaced, or obsoleted by other documents at any
   time. It is inappropriate to use Internet-Drafts as reference
   material or to cite them other than as "work in progress."

   The list of current Internet-Drafts can be accessed at
   http://www.ietf.org/1id-abstracts.html

   The list of Internet-Draft Shadow Directories can be accessed at
   http://www.ietf.org/shadow.html

Copyright Notice

   Copyright (C) The Internet Society (2001).  All Rights Reserved.

Abstract

   This document emphasizes the convergence of thinking regarding
   Internet naming in the industry and the technical community.

   There is strong consensus for establishing a layer above the current
   DNS to address some of its current limitations. At the same time, the
   community stresses the necessity for the new layer to go beyond the
   sole needs of the domain name system to realize an architecture
   capable of accommodating diverse services, with separate ownerships,
   different scopes and distinct operating models. In that context,
   interoperability is critical.

   This Internet-Draft introduces critical requirements for supporting
   multiple namespaces, in particular the crucial need for discovering
   namespaces across service providers. Acknowledging the direction



Arrouye, Parikh, and Popp                                       [Page 1]


draft-arrouye-kls-00.txt                                   1 August 2001


   opened by a reflection on the DNS [DNSROLE, DNSSEARCH] as well as the
   need to support flexible naming for various services [SLS], we
   describe the Keywords systems as a unique class of Service Lookup
   Systems [SLS].

1.  Introduction

   During the course of the past few years, a number of companies have
   looked into providing an easier user interface for Web navigation
   than that of typing a URI into a browser's address line. Examples of
   such companies are RealNames (worldwide network), Netword (North
   America), Netpia (Korea), and 3721 (China).

   All of these companies recognized the fact that URIs are unwieldy at
   best for users: they are hard to remember, due to their syntax, and
   as far as Web users are concerned, the only URIs that they commonly
   type into a browser are those forms that consist of the sole DNS name
   of the server they want to contact, eventually preceded by a valid
   protocol scheme. Therefore, identifiers like "www.ietf.org,"
   "www.yahoo.com," and others, have replaced URIs in daily interaction
   with the Web.

   Unfortunately, even these short names do not address users' needs.
   For example, while "www.ford.com" does belong to Ford Motor
   Corporation, it is "www.fordvehicles.com" that shows vehicles of the
   Ford brand. Not only is it unfortunate that people think of one name
   ("ford") and need to use another one ("fordvehicles"), but the DNS
   labels are not user-friendly names: "fordvehicles" has no space
   between the words, for example.

   While companies have been very creative in fitting their brand names
   into DNS labels (as can be seen by the company "Bird & Bird" getting
   the DNS name "twobirds.com"), if one looks at the Web in general,
   there is a big disconnect between the DNS labels that are used to
   identify Web sites, and the actual names by which people refer to
   these sites, which are typically brand or product names.

   This should not come as a surprise considering that the role of DNS
   is to map unique identifiers to network resources, not to give human
   friendly names to information or services. DNS has been abused to
   fulfill naming needs on the Web for a few years. It is actually a
   testimony to the strength of DNS that it has not crumbled under the
   additional requirements it has been put through.

   User behavior and expectations are just one part of the problem that
   DNS was not designed to and cannot be expected to address.  In
   addition, the Internet world is truly multilingual, bringing in a
   whole new set of names that are used to refer to information and



Arrouye, Parikh, and Popp                                       [Page 2]


draft-arrouye-kls-00.txt                                   1 August 2001


   services. People in different parts of the world want to use their
   own languages and scripts to refer to things, on the Web, in e-mail,
   on the Internet. So do companies. A Chinese company does not want to
   offer its Chinese products to its Chinese customers via a URI in
   Roman Script.

   Because DNS is the visible tip of the naming iceberg, the one that is
   already working and freely accessible (obtaining a domain name is
   pretty easy), people want to add capabilities to support DNS names
   that are truly multilingual, and the existence of the IETF
   Internationalized Domain Name (IDN) working group is a recognition of
   that desire.

   Unfortunately, a number of limitations of today's domain names still
   apply to IDNs as they are being considered now. Amongst them, the
   inability to use a space to separate words, strong limitations in the
   length of labels, and the lack to carry enough expression power to
   handle the visual requirements of names in Persian, for example.

   The technical community has begun to recognize that naming for humans
   is something that needs to be addressed outside of DNS, because DNS
   is a system of identifiers for network resources, not a system of
   easy-to-use names. Companies selling keywords and other friendly
   names have been the first to address this need.  One only needs to be
   aware of the existence of John Klensin's drafts about the role of DNS
   and how to extend it by adding other layers [DNSROLE, DNSSEARCH] to
   acknowledge that this view is becoming a well-regarded one.

   Keywords systems, which are based on natural language names in native
   languages, were the first to turn the recognition about the paradigm
   shift that happened in the use of DNS names into a new product and
   protocol layer.

   Their existence has fostered the creation of the Common Name
   Resolution Protocol [CNRP] in 1999. CNRP was an acknowledgment that
   users and applications wanted to use simple names to denote arbitrary
   resources on the Web. CNRP did not address the existence of many
   namespaces that were task-oriented.

   SLS builds on this foundation. It recognizes that people primarily
   use services such as the DNS, the Web, or email, and that those
   services use names to denote information. Using SLS, and thus CNRP,
   one can get from human-friendly names to network or information
   resource identifiers for a number of applications.

2.  Keyword Systems and Lessons Learned

   The primary goal of a keyword system is to make it easy for users to



Arrouye, Parikh, and Popp                                       [Page 3]


draft-arrouye-kls-00.txt                                   1 August 2001


   refer to things by their common, or human-friendly name in their
   local language. Keyword systems also take into account what users
   commonly expect - that the network will have information about them
   and use that context on their behalf to navigate them to the "best"
   destinations on the Web based on their query.

   It is a fact that a given name may refer to different things:
   "Woolworth" is a dime store in the USA but a high-end clothing store
   in South Africa; there may be one "Joe's Pizza" per city, or even
   more; and trademarks of the same name, but different applicability
   fields, coexist in a given country. Users, however, are typically
   only interested in giving one name to get some information.

   What RealNames learned from this is that while it is important for
   names to co-exist and be differentiated by facets(e.g. country, city,
   language, ...), it is also important that this differentiation is
   done transparently for the user wherever possible.

   Let's look at the RealNames keyword system for example. A Keyword has
   many facets, amongst them a name, a country, a language, a
   description, a URI, and a service code. Service codes (also called
   content codes) differentiate between the "web" and "mobile phone web"
   for which content is typically formatted differently. Countries let
   people refer to a global brand such as "Coca-Cola" by using the same
   name in different places. Language addresses the needs of countries
   with various ethnic populations, or official languages, to cater to
   the language requirements of their population. Description is a
   purely informative property. Finally, through a URI, one can point to
   the exact physical address associated to a given name in a given
   context.

   We've just used the word "context" and it is worth examining it. If
   names need to be associated to some meta-data to be useful and not
   fall into the flat namespace trap that the DNS has fallen into with
   gTLDs like .com, and, if users are going to expect satisfactory
   results using just names but not all the meta-data when using a
   keyword, then it is clear that the remaining meta-data must be
   supplied from the user's context.

   As namespaces will become more specialized, allowing for more
   identical names to exist in different dimensions, the reliance on
   context will increase. As we have operated our Keyword system these
   past four years, we have found that obtaining more meta-data is a
   difficult task. It is a formidable chicken and egg situation.  If
   users' applications and devices cannot supply the relevant data, then
   name buyers will not be willing to supply the appropriate ones when
   registering keywords. Keyword name buyers will only supply meta-data
   if they see a benefit. Benefits are derived from applications that



Arrouye, Parikh, and Popp                                       [Page 4]


draft-arrouye-kls-00.txt                                   1 August 2001


   perform more efficient navigation based on meta-data. No
   specification, regardless of how robust it may be, can alter this
   dynamic by itself. This combination of meta-data requirements and
   reluctance to provide meta-data cannot be ignored as we lay down the
   foundation for a new standard.

   At the same time, once a name buyer uses a Keyword system the
   importance of meta-data becomes clear and they are willing to provide
   appropriate meta-data to support the names they buy.

   However, name buyers are generally unwilling to supply more than a
   name at the time of use of the system (primarily because they take
   the rest of context for granted), so the burden of supplying the
   contextual information falls solely on the application or device.

   One key to resolving this problem is mining and using whatever
   context can be supplied. For example, as devices get smarter, they
   can also supply more context. For example, a Web browser typically
   knows about the user's language (from its language preferences, which
   then translate into an HTTP header for language negotiation), and can
   make a (risky) assumption about the country from the language. But a
   cellular phone nowadays can also know about the user's location to
   some reasonable accuracy. The existence of both devices also means
   that the kind of device is part of the available context.

   Another lesson from the requirement for context is that not all the
   properties of a keyword are useful in order to use the keyword as a
   way to go get some information (through the URI associated to it).
   The URI, which is what one wants to get, is definitely not one of
   these properties. Also, in a system where keywords are associated to
   a description, one does not want to require the description in order
   to enable direct resolution.

   In terms of supplying meta-data, the description associated with a
   keyword in the RealNames system is not necessary for lookups, while
   it is very useful for displaying entries in a directory of keywords.
   Some properties, like an industry category, are useful in both cases,
   but cannot be used today for navigation because of difficulty of
   establishing their value as part of a user's context.

   We call unique key the set of properties that are required to match a
   keyword and get a single result from a lookup. These properties
   translate into required context if an application wants to offer
   direct navigation from a name to a destination through a keywords
   system. Unique keys may differ from one keywords system to another
   (for example, RealNames uses a few properties while Netword only uses
   a name today). Unique keys may, and will, change, depending on the
   sophistication of the devices that are used to access a given



Arrouye, Parikh, and Popp                                       [Page 5]


draft-arrouye-kls-00.txt                                   1 August 2001


   service.

   As unique keys become smarter and able to know more about the context
   of their uses, more discriminating namespaces can come into existence
   and proliferate. Today, given the simplicity of the devices we use,
   keywords systems use short unique keys for the emerging namespaces
   that Keywords systems are.

   Because the unique key lets one access a record with other meta-data,
   it can be used not only to enable applications to fetch a URI and use
   it, but also to give one a place to store state.

3.  The Importance of Multiple Namespaces Above DNS

   A keyword system is an example of a namespace made up of records
   which consist of the unique keys and other helpful information. This
   namespace is dedicated to providing keyword-based natural language
   navigation. We believe that this is only one of many types of
   namespaces that will emerge as the Internet continues the significant
   transformation it is now undergoing.

   Like many in the industry, we believe that the limitations of DNS
   will be best addressed by facilitating the emergence of many
   interoperable namespaces. These multiple namespaces will be largely
   distributed, with distinct owners and administrative rules. In fact,
   that has already occurred, to the good of Internet users and service
   providers. We believe the right course is to encourage this
   innovation, and that industry and the technical community should work
   together to define a set of standard protocols through which all of
   these namespaces can interoperate.

   As the network continues its evolution, namespaces will offer
   different types of content and information to users and will be based
   on a continuum of business models. Although a namespace is likely to
   focus on one or a few types of network resources, it will be
   necessary for applications to interact with many namespaces at the
   same time in the course of a user session. Clearly, a flexible,
   lightweight protocol needs to define and facilitate interaction.

   This interaction is likely to encompass at the minimum registration,
   resolution (lookup) and discovery (search) services. Application
   developers must be able to rely on a common set of protocols for
   interacting with namespaces both to reduce the burden of adding new
   services to the network and also to enable users to freely choose
   between a rich offering of providers.

   Based on current industry development, it is not difficult to
   anticipate tomorrow's importance of companies, people, and products



Arrouye, Parikh, and Popp                                       [Page 6]


draft-arrouye-kls-00.txt                                   1 August 2001


   namespaces whose services will be delivered to user applications
   through the network and across multiple devices ranging from desktops
   to PDAs and data enabled cell phones.

   Many important instances of namespaces have already appeared in the
   last few years, and the fast-growing concept of Web services is
   likely to increase the reliance of user application on these
   namespaces as a core component. The increased tie and reliance
   between applications and namespaces accessible through the network
   exacerbates the need for open standards and interoperability.

   For example, if ICQ was one of the first major community namespace,
   Firefly was the first company to identify the need for separate
   specialized namespaces. Passport and its Hailstorm incarnation will
   be a namespace for people as network end-points that will not only
   provide unique identity (email address) to people but also personal
   context storage, and presence detection. Namespaces for people as
   network end-points are likely to become fundamental elements of all
   user-centric applications. Namespaces will likely encompass many
   aspects of our user experience.  Napster was the first P2P music
   namespace to profoundly impact the way users lookup and discover
   songs on the network and MusicNet is a more recent attempt to create
   a commercial music namespace that can be distributed through multiple
   applications.

   Other namespaces will be business enablers. UDDI is an industry-wide
   attempt to create a namespace for small and medium companies to
   discover each other and conduct business across the Internet and
   Keywords systems like RealNames are specialized toward human-friendly
   access to companies, products and services. There will obviously be
   many more valuable namespaces in the future.

   Since one cannot and does not want to anticipate any single one of
   them to dominate, it is important to facilitate the interaction of
   applications with multiple namespaces through a common sets of
   protocols accessible across multiple devices and operating systems.

4.  Requirements Arising From Multiple Independent Namespaces

   The growing proliferation of many independent namespaces places
   important constraints and requirements on the standard that will be
   created. We believe that some of these requirements may have been
   absent or minimized in previous proposals. Such critical constraints
   include:

   - The definition of a mechanism for discovering Namespaces.

   - The standardization of all core namespace services: resolution



Arrouye, Parikh, and Popp                                       [Page 7]


draft-arrouye-kls-00.txt                                   1 August 2001


   (lookup), discovery (search) and registration.

   - The need to simplify Klensin's layer 2 [DNSROLE] in order to
   facilitate the adoption of the standard by most client applications
   and service providers.

   - The support for different rules of uniqueness and disambiguation
   across namespaces.

   The first requirement is about enabling applications to discover all
   available Namespaces.

   The second is about service interoperability. In a world of many
   namespaces, these requirements are straightforward (and the solutions
   non trivial).

   The third requirement underlines the need to minimize the definition
   of the schema that Klensin's layer 2 services must all implement for
   interoperability. It is mainly driven by the need to simplify client
   implementations. As we impose more facets on this layer, we impose
   more burden on application developers and the user-interface
   (especially considering that as pointed out above, the user context
   carried by applications today is fairly minimal). This model cannot
   be successful in practice.

   Application developers' adoption of the standard is obviously
   essential since the prospect of increased distribution will be the
   main driver for namespace owners to implement the standard and make
   their service interoperable.

   However, we also recognize that from a service provider standpoint,
   the complexity of the Klensin's layer 2 schema is not a major
   impediment since facets within a query can simply be ignored. For
   example, all things being equal, a service that would not support
   category, for example, could return the same results whatever value
   of the category facet the user may specify. Obviously, one would
   expect that the ability to accurately match results to the query
   context passed by the client to be a strong service differentiator,
   hence a market force to drive providers to fully implement the
   required schema.

   This last observation leads to the third point. Like the authors of
   [SLS], we do not think that it is possible to impose a single context
   of uniqueness to all namespaces (we define uniqueness of a namespace
   as the minimum set of facets required to uniquely identify a record
   within the namespace). We too, believe that the proper notion of
   uniqueness is tied to the problem space tackled by the namespace, and
   hence, will vary from one provider to another.



Arrouye, Parikh, and Popp                                       [Page 8]


draft-arrouye-kls-00.txt                                   1 August 2001


   At the same time, we do not believe that it is possible to define a
   Klensin's layer 2 schema capable of providing all namespaces with
   sufficient context to ensure uniqueness and disambiguation for
   lookups.

   Instead, we see a practical solution in the middle. We believe that
   some namespaces will require narrower context of uniqueness than the
   complete schema that they expose to an application, using the extra
   facets to facilitate disambiguation by end-users through an
   interactive process (e.g. a user typing "Joe" could be asked to
   disambiguate between the unique Keyword "Joe's seafood" in the
   restaurant category and the unique Keyword "Joe body repair shop" in
   the "body shop" category).

   If this is true, defining the aforementionned layer as the ceiling to
   provide disambiguation for lookup is not very useful. Indeed, it is
   easy to envision that whichever standard set of facets we eventually
   agree upon for this layer, a new namespace will emerge with new
   requirements for uniqueness and disambiguation to prove us wrong.

   In that case, we propose that it might be more appropriate to
   consider defining the sets of facets that all namespaces should
   support in order to provide interoperability, while at the same time
   enabling each namespace to advertise to the applications the exact
   set of facets that it really needs for uniqueness. The definition of
   facets should be consistent, but each application should be able to
   dynamically choose which facets to use. As applications exist and
   will emerge that have different needs than DNS envisioned, use of
   facets should be determinable by each application, not by a protocol
   designed to suit the needs of the existing DNS system.

   To illustrate that last point, let us consider a music namespace that
   supports all the layer 2 facets proposed by John Klensin and a new
   one called "singer" (the common name in this namespace is the title
   of the song). Let us assume now that in this database of songs, the
   context provided by the song and the singer's name is enough to
   pinpoint a unique record. It is easy to see that:

   - It would not be sufficient for an application to only use the
   standardized layer 2 facets to disambiguate between songs (the
   "singer" property is missing).

   - It would be bad to prompt the user for the entire context
   (language, geography, and category) supported by the namespace for
   accessing a specific song, considering that in this case, only 2
   facets would suffice.

   - It would not be efficient for the application to store the full



Arrouye, Parikh, and Popp                                       [Page 9]


draft-arrouye-kls-00.txt                                   1 August 2001


   context of the record if the goal is to bookmark a song for future
   reference.

   This observation leads to separating the notion of schema for
   interoperability (the Klensin's layer 2 schema) from the notion of
   uniqueness. There seems to be an agreement in the technical community
   that one of the main reasons for introducing a new layer above DNS is
   the lack of support for context in DNS names. At the same time, there
   is no clear consensus on what scope of uniqueness should be
   standardized.

   Our feeling is that agreeing on a fixed context for uniqueness would
   be an error. Context will always vary across providers and
   namespaces, and it is not, nor should it be, a requirement to ensure
   interoperability.

   However, if we agree to let service providers disagree on rules of
   uniqueness, we believe it is still important for applications to be
   able to discover these rules at the service level. In fact,
   applications need to understand how much context needs to be captured
   from the user in order to disambiguate a query (lookup), and how much
   context needs to be saved to be able to reference the same unique
   record (bookmark). Without such capability, applications would have
   to default to using layer one identifiers for uniquely referencing a
   record.

   Hence, there is a need for formalizing the notion of context-based
   identifiers. Context-based identifiers or unique keys (to use a
   database terminology) would allow a namespace to publish its own
   context of uniqueness (the sets of facets that it uses for uniquely
   identifying a record). Service providers would advertise one or many
   of these unique keys.

   Unique keys could subsequently be recognized and used by applications
   to uniquely reference a record. We anticipate that applications will
   take advantage of unique keys to minimize the user interface and the
   interaction with the user for disambiguation during the lookup
   process. As explained above, in a keyword system, such interaction is
   kept to a minimum. The user simply types the common name (keyword)
   although the application transparently passes the required context
   (user country, language and the target service) for disambiguation.
   If the name is unambiguous, the user accesses the resource directly;
   otherwise a list of records is returned and presented to the user for
   an interactive disambiguation.

5.  Unique Keys

   To formalize the notion of a context-based identifier or unique key



Arrouye, Parikh, and Popp                                      [Page 10]


draft-arrouye-kls-00.txt                                   1 August 2001


   and enable each namespace to publish its rules of uniqueness, we
   propose to introduce to CNRP a new mechanism to declare which facets
   are part of the unique key.  This mechanism will be used to express
   the set of facets necessary to uniquely identify a single record for
   a given service (possibly within each different service type). In
   CNRP, a reference to facets is expressed using the
   "propertyreference" property. This makes it relatively
   straightforward for a service to describe its unique keys.

   To illustrate this last point, let us consider a namespace of people
   uniquely identified by their telephone number very much like the type
   of service provided by ENUM [RFC2916]. Such service could advertise
   its unique identifier comprised of the telephone number and the SLS
   service target as follows (the common name is the telephone number):

     <?xml version="1.0"?>
       <!DOCTYPE cnrp PUBLIC "-//IETF//DTD CNRP 1.0//EN"
        "http://ietf.org/dtd/cnrp-1.0.dtd">
       <cnrp>
         <results>
           <service>
             <serviceuri>urn:foo:bar</serviceuri>
             <servers>
               <server>
                 <serveruri>http://enum.example.com:4321</serveruri>
               </server>
             </servers>
             <description>This is the definition of an ENUM like
               SLS service</description>
             <!-- This schema defines the service as SLS compliant. -->
             <property name="cnrp-service-type">sls</property>
             <!-- We define all the properties of a keyword system.  -->
             <propertyschema>
               <propertydeclaration id="i1">
                 <propertyname>commonname</propertyname>
                 <propertytype>E.164</propertytype>
               </propertydeclaration>
               <propertydeclaration id="i2">
                 <propertyname>service</propertyname>
                 <propertytype>sls</propertytype>
               </propertydeclaration>
               <propertydeclaration id="i3">
                 <propertyname>userpublicprofile</propertyname>
                 <propertytype>freeform</propertytype>
               </propertydeclaration>
             </propertyschema>
             <!-- This is where we define the scope of uniqueness. -->
             <uniquekey>



Arrouye, Parikh, and Popp                                      [Page 11]


draft-arrouye-kls-00.txt                                   1 August 2001


               <propertyreference required="yes" ref="i1" />
               <propertyreference required="yes" ref="i2" />
             </uniquekey>
             <!-- We define all the properties in the query. -->
             <queryschema>
               <propertyreference required="yes" ref="i1" />
               <propertyreference required="yes" ref="i2" />
             </queryschema>
             <!-- We define all the properties in the response. -->
             <resourcedescriptorschema>
               <propertyreference required="yes" ref="i1" />
               <propertyreference required="yes" ref="i2" />
               <propertyreference required="no" ref="i3" />
             </resourcedescriptorschema >
           </service>
         </results>
       </cnrp>

   To put things in prospective, a typical query-response interaction
   between an application and the service would look like this (in this
   example, the application is looking for the user's email address):

     C:<?xml version="1.0"?>
     C:  <!DOCTYPE cnrp PUBLIC "-//IETF//DTD CNRP 1.0//EN"
     C:    "http://ietf.org/dtd/cnrp-1.0.dtd">
     C:  <cnrp>
     C:    <query>
     C:      <commonname>+330493656172</commonname>
     C:      <property name="service" type="sls">email</property>
     C:    </query>
     C:  </cnrp>

     S:<?xml version="1.0"?>
     S:  <!DOCTYPE cnrp PUBLIC "-//IETF//DTD CNRP 1.0//EN"
     S:    "http://ietf.org/dtd/cnrp-1.0.dtd">
     S:  <cnrp>
     S:    <results>
     S:      <service id="i0">
     S:        <serviceuri>http://enum.example.com</serviceuri>
     S:      </service>
     S:      <resourcedescriptor>
     S:         <commonname>+330493656172</commonname>
     S:         <resourceuri>mailto://jd@compuserve.com</resourceuri>
     S:         <property name="service"   type="sls">email</property>
     S:         <property name="userpublicprofile" type="freeform">This is
     S:           the best email address for Jean Dupont, Pediatrician
     S:           in Nice, France</property>
     S:      </resourcedescriptor>



Arrouye, Parikh, and Popp                                      [Page 12]


draft-arrouye-kls-00.txt                                   1 August 2001


     S:      </results>
     S:   </cnrp>

6.  Standardization of Keyword Systems as a Class of Namespaces

   This section describes a keyword system as a specific class of
   namespaces. Although we use CNRP and the SLS proposal to formalize
   the definition of a keyword system as a service type, the main goal
   is to identify the elements of a namespace that require
   standardization in order to define an interoperable class (aside from
   the requirements identified previously).

   Initiated by the CNRP effort, the standardization of keyword system
   is a goal that has been revived by the Multilingual Internet Name
   Consortium due to the rapid emergence of competing yet non-
   interoperable services in Asia, notably in Korea and China. We hope
   that this approach can establish a path toward the standardization of
   keyword systems that would proceed in conjunction with the definition
   of the broader directory layer above DNS.

   We feel that the standardization of keyword system must imply the
   following:

   1. The definition of a list of service targets relevant to all
   keyword systems (e.g. web, email, mobile-web, etc...)

   2. For each of these target services, the definition of the unique
   key common to all keyword systems.

   For example, the standard could stipulate that the unique key for
   keyword systems for the "web" target service be the combination of a
   common name, a language, and a country. On the other hand, for the
   target service "email", it is highly conceivable that the standard
   would be defined on a very different set of facets (for example,
   "organization name" could be one of the required facets).

7.  Formalization of KLS Using The SLS Notation

   Using CNRP, a client application could discover that a CNRP service
   is a keyword system by issuing the standard CNRP "servicequery":

     <?xml version="1.0"?>
       <!DOCTYPE cnrp PUBLIC "-//IETF//DTD CNRP 1.0//EN"
         "http://ietf.org/dtd/cnrp-1.0.dtd">
       <cnrp>
         <servicequery />
       </cnrp>




Arrouye, Parikh, and Popp                                      [Page 13]


draft-arrouye-kls-00.txt                                   1 August 2001


   The service would then return a service object expressing the
   characteristics common to all keyword systems (also note the
   suggested use of the cnrp-service-type property as defined in the SLS
   proposal):

     <?xml version="1.0"?>
       <!DOCTYPE cnrp PUBLIC "-//IETF//DTD CNRP 1.0//EN"
         "http://ietf.org/dtd/cnrp-1.0.dtd">
       <cnrp>
         <results>
           <service>
             <serviceuri>urn:foo:bar</serviceuri>
             <servers>
               <server>
                 <serveruri>http://keywords.ex.com:4321</serveruri>
               </server>
             </servers>
             <description>This is the definition of a keyword system as
               an SLS service</description>
             <!-- We define the service as an SLS service. -->
             <property name="cnrp-service-type">sls</property>
             <!-- We also define the service as a keyword system.  -->
             <property name="cnrp-service-type">keywords</property>
             <!-- We define all the properties of this keyword system. -->
             <propertyschema>
               <propertydeclaration id="i0">
                 <propertyname>commonname</propertyname>
                 <propertytype>freeform</propertytype>
               </propertydeclaration>
               <propertydeclaration id="i1">
                 <propertyname>language</propertyname>
                 <propertytype>RFC1766</propertytype>
               </propertydeclaration>
               <propertydeclaration id="i2">
                 <propertyname>geography</propertyname>
                 <propertytype>ISO3166-1</propertytype>
               </propertydeclaration>
               <propertydeclaration id="i3">
                 <propertyname>service</propertyname>
                 <propertytype>sls</propertytype>
               </propertydeclaration>
             </propertyschema>
             <!-- We define all the properties admissible and required
                  in the query. -->
             <queryschema>
               <propertyreference required="yes" ref="i0" />
                 <propertyreference required="yes" ref="i1" />
                 <propertyreference required="yes" ref="i2" />



Arrouye, Parikh, and Popp                                      [Page 14]


draft-arrouye-kls-00.txt                                   1 August 2001


               <propertyreference required="yes" ref="i3" />
             </queryschema>
             <!-- We define all the properties admissible and
                  required in the response. -->
             <resourcedescriptorschema>
               <propertyreference required="yes" ref="i0" />
               <propertyreference required="yes" ref="i1" />
               <propertyreference required="yes" ref="i2" />
               <propertyreference required="yes" ref="i3" />
             </resourcedescriptorschema >
             <!-- This is where we define the scope of uniqueness. -->
             <uniquekey>
               <property name="service" type="sls">web</property>
               <propertyreference required="yes" ref="i0" />
               <propertyreference required="yes" ref="i1" />
               <propertyreference required="yes" ref="i2" />
               <propertyreference required="yes" ref="i3" />
             </uniquekey>
           </service>
         </results>
       </cnrp>

   Note that the queryschema is only one path for an application to
   discover that a CNRP service is a keyword system. Since the service
   object is also embedded within the response to any query, the
   application can simply recognize a keyword system by parsing the
   information contained in the service object before processing the
   query results.

8.  Conclusion

   As the network transforms, meta-data that provides context is
   critical, and a flexible set of standards regarding collection and
   use of meta-data must be defined. This will allow a wide
   proliferation of namespaces to quickly develop that meet the needs of
   the diverse worldwide population that uses the Internet.


9.  References

   [CNRP] N. Popp, M. Mealling, and M. Moseley, Common Name Resolution
   Protocol (CNRP), draft-ietf-cnrp-10.txt, June 2001.

   [DNSROLE] J. Klensin, Role of the Domain Name System, draft-klensin-
   dns-role-01.txt, May 2001.

   [DNSSEARCH] J. Klensin, A Search-based access model for the DNS,
   draft-klensin-dns-search-01.txt, July 2001.



Arrouye, Parikh, and Popp                                      [Page 15]


draft-arrouye-kls-00.txt                                   1 August 2001


   [RFC2916] P. Faltstrom, E.164 number and DNS, RFC 2916, September
   2000.

   [SLS] M. Mealling and L. Daigle, Service Lookup System (SLS), draft-
   mealling-sls-00.txt, July 2001.

Author's Address

   Yves Arrouye
   RealNames Corporation
   150 Shoreline Drive
   Redwood City, CA 94065

   Phone: (650) 486-5503

   E-mail: yves@realnames.com

   Keyword: RealNames
   Web: http://www.realnames.com/

   Vishesh Parikh
   RealNames Corporation
   150 Shoreline Drive
   Redwood City, CA 94065

   Phone: (650) 486-5507

   E-mail: vparikh@realnames.com

   Keyword: RealNames
   Web: http://www.realnames.com/

   Nico Popp
   RealNames Corporation
   150 Shoreline Drive
   Redwood City, CA 94065

   Phone: (650) 486-5549

   E-mail: nico@realnames.com

   Keyword: RealNames
   Web: http://www.realnames.com/

Full Copyright Statement

   Copyright (C) The Internet Society (2001).  All Rights Reserved.




Arrouye, Parikh, and Popp                                      [Page 16]


draft-arrouye-kls-00.txt                                   1 August 2001


   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.























Arrouye, Parikh, and Popp                                      [Page 17]


Html markup produced by rfcmarkup 1.129d, available from https://tools.ietf.org/tools/rfcmarkup/