Internet Draft                                          Karen R. Sollins
draft-ietf-urn-req-frame-02.txt                                  MIT/LCS
Expires December 4, 1997                                    June 4, 1997

        Guidelines and a Framework for URN Resolution Systems


Status of this draft
     This document is an Internet-Draft.  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.''

     To learn the current status of any Internet-Draft, please check
     the ``1id-abstracts.txt'' listing contained in the Internet-
     Drafts Shadow Directories on ftp.is.co.za (Africa),
     nic.nordu.net (Europe), munnari.oz.au (Pacific Rim),
     ds.internic.net (US East Coast), or ftp.isi.edu (US West Coast).


Abstract:

This document addresses the issues of the discovery of URN resolver
services that in turn will directly translate URNs into URLs and URCs.
The document falls into three major parts, the assumptions underlying
the work, the guidelines in order to be a viable Resolver Discovery
Service or RDS to help in finding URN resolvers, and a framework for
designing RDSs.  The guidelines fall into three major areas:
evolvability, usability, and security and privacy.  An RDS that is
compliant with the framework will not necessarily be compliant with the
guidelines.  Compliance with the guidelines will need to be validated
separately.


. Introduction

The purpose of this document is to lay out the engineering criteria for
what we will call here a Resolver Discovery Service (RDS), a service to
help in the learning about URN resolvers.  This is a component of the
realization of an information infrastructure.  In the case of this work,
that infrastructure is to be available, "in the Internet" or globally,
and hence the solutions to the problems we are addressing must globally
scalable.  In this work, we are focussing specifically on naming of
resources and resolution of those names to the exclusion of other
problems such as typing, resource access and availability, security of
the resources, etc.  Those are all important problems, but not part of
this effort.

The Uniform Resource Identifier Working Group defined a naming
architecture, as demonstrated in a series of three RFCs 1736[RFC1736],

                               - 1 -

1737{RFC1737}, and 1738[RFC1738].  Although several further documents
are needed to complete the description of that architecture, it
incorporates three core functions often associated with "naming":
identification, location, and mnemonics or semantics.  By location, we
mean fully-qualified Domain Names or IP addresses.  Names may provide
the ability to distinguish one resource from another, by distinguishing
their "names".  Names may help to provide access to a resource by
including "location" information.  Lastly, names may have other semantic
or mnemonic information that either helps human users remember or figure
out the names, or include other semantic information about the resource
being named.  The URI working group concluded that there was need for
persistent, globally unique identifiers, distinct from location or other
semantic information; these "names" provide identity, in that if two of
them are "the same" (under some simple rule of canonicalization), they
identify the same resource.  Furthermore, the group decided that these
"names" were generally to be for machine, rather than human,
consumption.  One can imagine a variety human-friendly naming (HFN)
schemes supporting different suites of applications and user
communities.  These will need to provide mappings to URNs in tighter or
looser couplings, depending on the namespace.  It is these HFNs that
will be mnemonic, content-full, and perhaps mutable, to track changes in
use and semantics.  They may provide nicknaming and other aliasing,
relative or short names, context sensitive names, descriptive names,
etc.  The URI naming architecture as described in the introductions to
RFCs 1736 and 1737 lays out three sorts of components to the naming
architecture: identifiers called Uniform Resource Names (URNs), locators
called Uniform Resource Locators (URLs) and semantic meta-information
called Uniform Resource Characteristics (URCs).  This document focusses
on part of the problem of the translation from URN to URL and/or URC.

URNs as described in RFC 1737 are defined globally; they are ubiquitous
in that a URN anywhere in any context identifies the same resource.
With respect to an RDS we must ask what URN ubiquity implies for the RDS
and for resolution in general.  But in terms of Internet services and
accessibility, there can be no systematic guarantees.  In addition, it
is quite possible that the resolution of a URN to an instance of a
resource may reach different instances or copies under different
conditions.  Thus, although a URN anywhere refers to the same resource,
in some locations under some conditions, and at different times, due to
either the vagueries of network conditions or policy controls a URN may
sometimes be resolvable and other times or places not.  Ubiquitous
resolution cannot be assumed simply because naming is ubiquitous.  On
the other hand wide deployment and usage will be an important feature of
any RDS design.

Within the URI community there has been a concept used frequently that
for lack of a better term we will call a _hint_.  A hint is something
that helps in the resolution of a URN; we map URNs to hints as an
interim stage in accessing a resource.  A hint may also have
meta-information associated with it, such as an expiration_time or
certification of authenticity.  We expect that these will stay with a
hint rather than being managed elsewhere.  We will assume in all further
discussion of hints that they include any necessary meta-information as
well as the hint information itself.  Examples of hints are: 1) the name
of a resolver service that may further resolve the URN, 2) the address
of such a service, 3) a location at which the resource was previously

                               - 2 -

found.  The defining feature of hints is that they are only hints; they
may be out of date, temporarily invalid, or only applicable within a
specific locality.  They do not provide a guarantee of access, but they
probably will help in the resolution process.  We must assume that most
resolutions of URNs will be provided by the use of locally stored hints,
because maintaining a database of globally available, completely
up-to-date location information is infeasible for performance reasons.
There are a number of circumstances in which one can imagine that hints
become invalid, either because a resource has moved or because a
different URN resolver service has taken over the responsibility for
resolution of the URN.  Hints may be found in a variety of places.  It
is generally assumed that a well engineered system will maintain a set
of hints for each URN at each location where that URN is found.  In
addition, for those situations in which those hints found locally fail,
a well-engineered system will provide a fall-back mechanism for
discovering further hints.  It is this fall-back mechanism, an RDS, that
is being addressed in this document.  As with all hints, there can never
be a guarantee that access to a resource will be available to all
clients, even if the resource is accessible to some.  However, an RDS is
expected to work with reasonably high reliability, and, hence, may
result in increased response time.

The remainder of this document falls into three sections.  The first
identifies several sets of assumptions underlying this work.  There are
three general assumptions:
   * URNs are persistant;
   * URN assignment can be delegated;
   * Decisions can be made independently, enabling isolation from decisions
     of one's peers.

The next section lays out the guidelines for a Resolver Discovery
Service.  This section is probably the most critical of the document,
because it is this that provides the metric for whether or not a
proposed scheme for a n RDS is adequate or not.  To summarize, there are
three core rubrics, each of which is refined and subdivided below:
   R1) An RDS must allow for evolution and evolvability;
   R2) Usability of an RDS with regard to each of the sets of constituents
       involved in the identification and location or resources is paramount;
   R3) It is centrally important that the security and privacy needs of all
       consituents be feasibly supported, to the degree possible.

It is important to note that the origins of this document were as a
requirements document.  Therefore it retains its flavor of a requirments
documents including the use of "must" and "should".  The consensus of
the working group currently is that more experience is needed before it
can have the confidence necessary to be explicit about requirements for
RDSs.  Hence the document is worded in terms of "guidelines" and
"rubrics", with the understanding that anyone any proposal for an RDS
design should still measure up to the statements in this document, based
on the accrued knowledge and experience of a group that has been working
in this area for a number of years.  Any RDS proposal should document
how it addresses each of the rubrics.  If it does not adequately address
any of them, it should document the reasoning behind it, so that the
community can learn from that experience, with the intention of defining
a set of requirements in the future.


                               - 3 -

For the reader short on time, each of the three major subsections of the
guidelines section begins with a summary list of the more detailed
guidelines identified in that section.  The final section of the
document lays out a framework for such RDSs.  The purpose of this last
section is to bound the search space for RDS schemes.  One must be
careful not to assume that because an RDS scheme fits within the
framework that it necessarily meets the guidelines.  As will be
discussed further in this last section, designing within the framework
does not guarantee compliance, so compliance evaluation must also be
part of the process of evaluation of a scheme.

. Assumptions

Based on previous internet drafts and discussion in both the URN BOFs
and on the URN WG mailing list, three major areas of assumptions are
apparent: longevity, delegation, and independence.  Each will be
discussed separately.

The URN guidelines state that a URN is to be a "persistent identifier".
It is probably the case that nothing will last forever, but in the time
frame of resources, users of those resources, and the systems to support
the resources, the identifier should be considered to be persistent or
have a longer lifetime than those other entities.  There are two
assumptions that are implied by longevity of URNs: mobility and
evolution.  "Mobility" assumes that everything will move over the life
of a URN.  For example, resources will move from one machine to another,
because individual machines have a much shorter lifetime than resources,
generally measured in a number of years less than a decade.  Owners of
resources may move and wish their resources to follow them.  The
services themselves will move.  "Evolution" assumes that the supporting
infrastructure will evolve.  This may take the form of entirely new
transport protocols or new versions of existing protocols.  Furthermore,
services such as storage services may evolve; it is even possible that
within a human lifetime the Unix file system model may no longer be in
use!  Clearly there will be evolution of and improvement in supporting
authentication and security mechanisms.  These are only examples.  In
general, we must assume that almost any piece of the supporting
infrastructure of URN resolution will evolve.  In order to deal with
both the mobility and evolution assumptions that derive from the
assumption of longevity, we must assume that users and their
applications can remain independent of these mutating details of the
supporting infrastructure.

The second and third assumptions are two forms of modularity: delegation
and isolation.  The delegation assumption is that an entity may
partition and pass off some of its authority or responsibility.  One of
those responsibilities is for assigning URNs; practically speaking,
there cannot be only a single authority for assigning URNs.  We expect
that there will be a multi-tiered naming authority delegation.
Furthermore, it is difficult to imagine a non-partitioned and delegated
global RDS, meaning that hint discovery and resolution will be
partitioned and delegated.  In some RDS schemes, the delegation of
naming authority will form a basis for delegating the management and
dispensing of location information.


                               - 4 -

The third assumption is independence or isolation of one authority from
another and, at least to some extent from its parent.  Underlying much
of the thinking and discussion in the URI and URN working groups has
been the assumption that when a component delegates authority to another
component, the delegatee can operate in that domain independently of its
peers and within bounds specified by the delegation, independently of
the delegator.  This isolation is critically important in order to allow
for independence of policy and mechanism.

There are a number of more specific assumptions that fall under this
rubric of isolation.  First, we assume that the publisher of a resource
can choose resolver services, independently of choices made by others.
At any given time, the owner of a namespace may choose a particular URN
resolver service for that delegated namespace.  Such a URN resolver
service may be outside the RDS service model, and just identified or
located by the RDS service.  Second, it must be possible to make a
choice among RDS services, perhaps based on different underlying
internal architectures.  The reason that this is an assumption is that
there must be an evolutionary path through a sequence of core RDS
services.  Although at any given time there is likely to be only one or
a small set of such services, the number is likely to increase during a
transition period from one architecture to another.  Thus, it must be
assumed that clients can make a choice among a probably very small set
of RDSs.  Third, there must be independence in the choice about levels
and models of security and authenticity required.  This choice may be
made by the owner of a naming subspace, in controlling who can modify
hints in that subspace.  A naming authority may delegate this choice to
the owners of the resources named by the names it has assigned.  There
may be limitations on this freedom of choice in order to allow other
participants to have the level of security and authenticity they
require, for example, in order to maintain the integrity of the RDS
infrastructure as a whole.  Fourth, there is an assumption of
independence of choice of the rule of canonicalization of URNs within a
namespace, limited by any restrictions or constraints that may have been
set by its parent namespace.  This is a choice held by naming
authorities over their own subnamespaces.  Rules for canonicalization
will be discussed further in the framework section below.  Thus, there
are assumptions of independence and isolation to allow for delegated,
independent authority in a variety of domains.

The modularity assumptions of delegation and isolation imply
independence of decision and implementation, leading to a
decentralization that provides a certain degree of safety from denial of
service.  Based on these these assumptions in conjunction with that of
longevity and those for URLs and URNs as detailed in RFCs 1736 and 1737,
we can now turn to the guidelines for a Resolver Discovery Service.

. Guidelines

The guidelines or rubrics applying to a Resolver Discovery Service or
RDS center around three important design goals: evolvability, usability,
and security and privacy.  At its core the function of an RDS is to
provide hints for accessing a resource given a URN for it.  These hints
may range in applicability from local to global, and from short-lived to
long-lived.  They also may vary in their degree of verifiable
authenticity.  While it may be neither feasible nor necessary that

                               - 5 -

initial implementations support every guideline, every implementation
must support evolution to systems that do support every rubric.

It is important to note that there are requirements, not applicable
specifically to an RDS that must also be met.  A whole URN system will
consist of namespaces, the resolution information for them, and the
mapping from names in the namespaces to resolution information (or
hints).  URN schemes must meet the requirements of RFC 1737.  Resolution
information, to the extent it is expressed as URLs must meet the
requirements of RFC 1736.  But this does not tell the whole story.
Although the URN working group will identify several acceptable
namespaces and the rules binding them, such as how delegation occurs,
how it is expressed in the names, how and to what extent binding to hint
information will be constrained by the namespace, in the long run a
document will be needed to guide the evaluation criteria for acceptance
of new namespaces.  These are not included in the list of guidelines
below because they are not guidelines or requirements for an RDS, but
rather are requirements for naming schemes themselves.

Each section below begins with a summary of the points made and
discussed in the following discussion.  It is worth noting here that
there is some degree of overlap among the areas, such as in allowing for
the evolution of security mechanisms, etc.  Issues may appear in more
than one place.  It is also important to recognize that conformance with
the rubrics may often be subjective.  Most of these rubrics are not
quantifiable and hence conformance is a judgment call and a matter of
degree.  Lastly, the reader may find that some of them are those of
general applicability to distributed systems and some are specific to
URN resolution.  Those of general applicability are included for
completeness and are not distinguished as such.

3.1 Evolution

The issues in the area of evolvability are:

   R1.1) An RDS must be able to support scaling updwards both in terms
         of the number of resources for which URNs will be required and
         in terms of the number of publishers and users of those
         resources;

   R1.2) A hint resolution environment must support evolution of
         mechanisms, specifically for:
         * a growing set of URN schemes;
         * new kinds local URN resolver services;
         * new authentication schemes;
         * alternative RDS schemes active simultaneously;
   R1.3) An RDS must be capable of supporting the separation of global
         identification from location information;
   R1.4) An RDS must allow the development and deployment of
         administrative control mechanisms to manage human behavior with
         respect to limited resources.

One of the lessons of the Internet that we must incorporate into the
development of mechanisms for resolving URNs is that we must be prepared
for change.  Such changes may happen slowly enough to be considered
evolutionary modifications of existing services or dramatically enough

                               - 6 -

to be considered revolutionary.  They may permeate the Internet universe
bit by bit, living side by side with earlier services or they may take
the Internet by storm, causing an apparent complete transformation over
a short period of time.  There are several directions in which we can
predict the need for evolution, even at this time, prior to the
deployment of any such service.  At the very least, the community and
the mechanisms proposed should be prepared for these.

First, scaling is a primary issue in conjunction with evolution.  The
number of users and publishers is certainly on an increasing trajectory.
One might consider that it has an upper limit based on the population,
but that assumes that resources are only published by and for the use of
humans.  As our world becomes more automated, more "users" will be
electronic.  In addition, clearly the number of resources will grow by
orders of magnitude.  Hence the number of URNs will also increase
similarly.  These facts mean that an RDS design must be prepared to
handle increasing numbers of requests for inclusion, update and
resolution.  This is not to say that there will necessarily be more
updates or resolutions per URN; we cannot predict that at this time.
Any design is likely to perform less well above some set of limits, so
it is worth considering the growth limitations of each design
alternative.

Second, we expect there to be additions and changes to the mechanisms.
The community already understands that there must be a capacity for new
URN schemes.  A URN scheme will define a set of URNs that meet the URN
requirements[RFC1737], but may have further constraints on the internal
structure of the URN. The intention is that URN schemes can be free to
specify parts of the URN that are left opaque in the larger picture.  In
fact, a URN scheme may choose to make public the algorithms for any such
"opaque" part of the URN. For example, although it may be unnecessary to
know the structure of an ISBN, the algorithm for understanding the
structure of an ISBN has been made public.  Other schemes may either
choose not to make their algorithms public, or choose a scheme in which
knowledge of the scheme does not provide any significant semantics to
the user.  In any case, we must be prepared for a growing number of URN
schemes.

Often in conjunction with a new URN scheme, but possibly independently
of any particular URN scheme, new kinds of resolver services may evolve.
For example, one can imagine a specialized resolver service based on the
particular structure of ISBNs that improves the efficiency of finding
documents given their ISBNs.  Alternatively, one can also imagine a
general purpose resolver service that trades performance for generality;
although it exhibits only average performance resolving ISBNs, it makes
up for this weakness by understanding all existing URN schemes, so that
its clients can use the same service to resolve URNs regardless of
naming scheme.  In this context, there will always be room for
improvement of services, through improved performance, better
adaptability to new URN schemes, or lower cost, for example.  In any
case, new models for URN resolution will evolve and we must be prepared
to allow for their participation in the overall resolution of URNs.

If we begin with one overall plan for URN resolution, into which the
enhancements described above may fit, we must also be prepared for an
evolution in the authentication schemes that will be considered either

                               - 7 -

useful or necessary in the future.  There is no single globally accepted
authentication scheme, and there may never be one.  Even if one does
exist at some point in time, there will always be threats to it, and so
we must always be prepared to move on to newer and better schemes, as
the old ones become too easily spoofed or guessed.

Lastly, in terms of mechanism, although we may develop and deploy a
single RDS scheme initially, we must be prepared for that top level
model to evolve.  Thus, if the RDS model supports an apparently
centralized (from a policy standpoint) scheme for inserting and
modifying authoritative information, over time we must be prepared to
evolve to a different model, perhaps one that has a more distributed
model of authority and authenticity.  If the model has no core but
rather a cascaded partial discovery of information, we may find that
this becomes unmanageable with an increase in scaling.  Whatever the
core of the model, we must be prepared for it to evolve with changes in
scaling, performance, and policy constraints such as security and cost.

Second, in addition to the evolution of resolution mechanisms, we expect
that the community will follow an evolutionary path towards the
separation of location information from identification.  The URN
requirements document suggested this path as well, and there has been
general agreement in much of the community that such a separation is
desirable.  This is a problem that the public at large has generally not
understood.  Today we see the problem most clearly with the use of URLs
for identification.  When a web page moves, its URL becomes invalid.
Suppose such a URL is embedded in some page, stored in long term
storage.  There are three possible outcomes to this scenario.  One
possibility is that the client is left high and dry with some message
saying that the page cannot be found.  Alternatively, a "forwarding
pointer" may be left behind, in the form of an explicit page requesting
the client to click on a new URL. Although this will allow the client to
find the intended page, the broken link cannot be fixed because the URL
is embedded in a file outside of the client's control.  A third
alternative is that the target server supplies redirect so that the new
page is provided for the client automatically.  In this case, the client
may not even realize that the URL is no longer correct.  The real
problem with both of these latter two situations is that they only work
as long as the forwarding pointer can be found at the old URL. Location
information was embedded in the identifier, and the resolution system
was designed to depend on that location information being correct.
There are few cases in which we can expect such information to remain
valid for a long time, but in many cases references need to have long
lifespans.  Most documents are only useful while their references still
function.  To the extent that an RDS scheme supports the separation of
global identification from location information it will be encouraging
the longer utility of the identities.

A third evolutionary issue is even more mechanical than the others.  At
any point in time, the community is likely to be supporting a compromise
position with respect to resolution.  We will probably be operating in a
situation balanced between feasibility and the ideal, perhaps with
policy controls used to help stabilize use of the service.  Ideally, the
service would be providing exactly what the customers wanted and they in
turn would not request more support than they need.  Since we will
always be in a situation in which some service provision resources will

                               - 8 -

be in short supply, some form of policy controls will always be
necessary.  Some policy controls may be realized as mechanisms within
the servers or in the details of protocols, while others may only be
realized externally to the system.  For example, suppose hint entries
are being submitted in such volume that the hint servers are using up
their excess capacity and need more disk space.  An effective solution
to this problem would be a mechanism such as a pricing policy.  This
pricing policy has the dual effect of both encouraging conservative use
of resources and collecting revenue for the improvement and maintenance
of the system.  We can also imagine administrative policy controls with
the force of laws or other social pressures behind them, but with no
technical mechanism enforcing or enabling them.  As technology changes
and the balance of which resources are in short supply changes, the
mechanisms and policies for controlling their use must evolve as well.

3.2 Usability and Feature Set Issues

To summarize, the usability rubrics fall into three areas based on
participation in hint management and discovery:

   R2.1) The publisher
      R2.1.1) URN to hint resolution must be correct and efficient with
              very high probability;
      R2.1.2) Publishers must be able to select and move among URN
              resolver services to locate their resources;
      R2.1.3) Publishers must be able to arrange for multiple access
              points for their location information;
      R2.1.4) Publishers should be able to provide for both long-lived
              and short-lived hints;
      R2.1.5) It must be relatively easy for publishers to specify to
              the management and observe their hint information as well
              as any security constraints they need for their hints.
   R2.2) The client
      R2.2.1) The interface to the RDS must be simple, effective, and
              efficient;
      R2.2.2) The client and client applications must be able to
              understand the information stored in and provided by the
              RDS easily, in order to be able to make informed choices.
   R2.3) The management
      R2.3.1) The management of hints must be as unobtrusive as
              possible, avoiding using too many network resources;
      R2.3.2) The management of hints must allow for administrative
              controls that encourage certain sorts of behavior deemed
              necessary to meet other requirements;
      R2.3.3) The configuration and verification of configuration of
              individual RDS servers must be simple enough not to
              discourage configuration and verification.

Usability can be evaluated from three distinct perspectives: those of a
publisher wishing to make a piece of information public, those of a
client requesting URN resolution, and those of the provider or manager
of resolution information.  We will separately address the usability
issues from each of these three perspectives.

It is worth noting that there are two additional sorts of participants
in the whole naming process, as discussed in the URN WG.  They are the

                               - 9 -

naming authorities which choose and assign names, and the authors who
include URNs in their resources.  These two are not relevant to the
design of an RDS and hence are not discussed further here.

3.2.1 The Publisher

The publisher must be able to make URNs known to potential customers.
From the perspective of a publisher, it is of primary importance that
URNs be correctly and efficiently resolvable by potential clients with
very high probability.  Publishers stand to gain from long-lived URNs,
since they increase the chance that references continue to point to
their published resources.

The publisher must also be able to choose easily among a variety of
potential services that might translate URNs to location information.
In order to allow for this mobility among resolver services, the
architecture for resolver services specified within the IETF should not
result in a scenario in which changing from one resolver service to
another is an expensive operation.

The publisher must be able to arrange for multiple access points to a
published resource.  For this to be useful, resolver services should be
prepared to provide different resolution or hint information to
different clients, based on a variety of information including location
and the various access privileges the client might have.  For example,
companies might arrange for locally replicated copies of popular
resources, and would like to provide access to the local copies only for
their own employees.  This is distinct from access control on the
resource as a whole, and may be applied differently to different copies.

The publisher should be able to provide both long and short term
location information about accessing the resource.  Long term
information is likely to be such information as the long term or the
location or identity of a resolver service with which the publisher has
a long term relationship.  One can imagine that the arrangement with
such a long term "authoritative" resolver service might be a guarantee
of reliability, resiliency to failure, and atomic updates.  Shorter term
information is useful for short term changes in services or to avoid
short lived congestion or failure problems.  For example, if the actual
repository of the resource is temporarily inaccessible, the resource
might be made available from another repository.  This short term
information can be viewed as temporary refinements of the longer term
information, and as such should be more easily and quickly made
available, but may be less reliable.

Lastly, the publishers will be the source of much hint information that
will be stored and served by the manager of the infrastructure.  Despite
the fact that many publishers will not understand the details of the RDS
mechanism, it must be easy and straightforward for them to install hint
information.  The publisher must be able not only to express hints, but
also to verify that what is being served by the manager is correct.
Furthermore, to the extent that there are security constraints on hint
information, the publisher must be able to both express them and verify
compliance with them easily.


                               - 10 -

3.2.2 The Client

From the perspective of the client, simplicity and usability are
paramount.  Of critical importance to serving clients effectively is
that there be an efficient protocol through which the client can acquire
hint information.  Since resolving the name is only the first step on
the way to getting access to a resource, the amount of time spent on it
must be minimized.

Furthermore, it will be important to be able to build simple, standard
interfaces to the RDS so that both the client and applications on the
client's behalf can interpret hints and subsequently make informed
choices.  The client, perhaps with the assistance of the application,
must be able to specify preferences and priorities and then apply them.
If the ordering of hints is only partial, the client may become directly
involved in the choice and interpretation of them and hence they must be
understandable to that client.  On the other hand, in general it should
be possible to configure default preferences, with individual
preferences viewed as overriding any defaults.

From the client's perspective, although URNs will provide important
functionality, the client is most likely to interact directly only with
human friendly names (HFNs).  As in direct human interaction (not
computer mediated), the sharing of names will be on a small, private, or
domain specific scale.  HFNs will be the sorts of references and names
that are easy to remember, type, choose among, assign, etc.  There will
also need to be a number of mechanisms for mapping HFNs to URNs.  Such
services as "yellow pages" or "search tools" fall into this category.
Although we are mentioning HFNs here, it is important to recognize that
HFNs and the mappings from HFNs to URNs is and must remain a separate
functionality from an RDS.  Hence, although HFNs will be critical to
clients, they do not fall into the domain of this document.

3.2.3 The Management

Finally, we must address the usability concerns with respect to the
management of the hint infrastructure itself.  What we are terming
"management" is a service that is distinct from publishing; it is the
core of an RDS.  It involves the storage and provision of hints to the
clients, so that they can find published resources.  It also provides
security to the extent that there is a commitment for provision of such
security; this is addressed in Section 3.3 below.

The management of hints must be as unobtrusive as possible. First, its
infrastructure (hint storage servers and distribution protocols) must
have as little impact as possible on other network activities.  It must
be remembered that this is an auxiliary activity and must remain in the
background.

Second, in order to make hint management feasible, there may need to be
a system for administrative incentives and disincentives such as pricing
or legal restrictions.  Recovering the cost of running the system is
only one reason for levying charges.  The introduction of payments often
has a beneficial impact on social behavior.  It may be necessary to
discourage certain forms of behavior that when out of control have
serious negative impact on the whole community.  At the same time, any

                               - 11 -

administrative policies should encourage behavior that benefits the
community as a whole.  Thus, for example, a small one-time charge for
authoritatively storing a hint will encourage conservative use of hints.
If we assume that there is a fixed cost for managing a hint, then the
broader its applicability across the URN space, the more cost effective
it is.  That is, when one hint can serve for a whole collection of URNs,
there will be an incentive to submit one general hint over a large
number of more specific hints.  Similar policies can be instituted to
discourage the frequent changing of hints.  In these ways and others,
behavior benefitting the community as a whole can be encouraged.

Lastly, symmetric to issues of usability for publishers, it must also be
simple for the management to configure the mapping of URNs to hints.  It
must be easy both to understand the configuration and to verify that
configuration is correct.  With respect to management, this issue may
have an impact not only on the information itself but also on how it is
partitioned among network servers that collaboratively provide the
management service or RDS.  For example, it should be straightforward to
bring up a server and verify that the data it is managing is correct.
Although this is not a rubric, it is worth nothing that since we are
discussing a global and probably growing service, encouraging volunteer
participants suggests that, as with the DNS, such volunteers can feel
confident about the service they are providing and its benefit to both
themselves and the rest of the community.


3.3 Security and Privacy Issues

In summary, security and privacy rubrics can be identified as some
degree of protection from threats.  These rubrics are all stated in
terms of possibilities or options for users of the service to require
and utilize.  Hence they address the availability of functionality, but
not for the use of it.  We recognize that all security is a matter of
degree and compromise.  These may not satisfy all potential customers,
and there is no intention here to prevent the building of more secure
servers with more secure protocols to suit their needs.  These are
intended to satisfy the needs of the general public.

   R3.1) It must be possible to create authoritative versions of a hint
         with access-to-modification privileges controlled;
   R3.2) It must be possible to determine the identity of servers or
         avoid contact with unauthenticated servers;
   R3.3) It must be possible to reduce the threat of denial of service
         by broad distribution of information across servers.
   R3.4) It must be possible within the bounds of organization policy
         criteria to provide at least some degree of privacy for
         traffic.
   R3.5) It must be possible for publishers to keep private certain
         information such as an overall picture of the resources they
         are publishing and the identity of their clients;
   R3.6) It must be possible for publishers to be able to restrict
         access to the resolution of the URNs for the resources they
         publish, if they wish.

When one discusses security, one of the primary issues is an enumeration
of the threats being considered for mitigation.  The tradeoffs often

                               - 12 -

include cost in money and computational and communications resources,
ease of use, likelihood of use, and effectiveness of the mechanisms
proposed.  With this in mind, let us consider a set of threats.

A good place to begin is with the early work of Voydock and Kent [VK83].
They identify unauthorized release of information as a passive attack.
On the other hand, unauthorized modification of information, denial of
service, and spurious association initiation are labelled as active
attacks.  An intruder at any protocol layer can attack at any of the
links or computational elements (hosts, routers, etc.)  at that layer.
Attacks at one layer can be achieved by subverting or attacking the
lower layers.  An unauthorized release of information is a violation of
privacy or confidentiality.  This may be achieved by a release of the
information itself.  Additional passive threats are from secondary
information through traffic analysis or other violations of transmission
security, such as noticing lengths and/or sources and destinations of
traffic.  Moving to the active threats, unauthorized modification of
information can be partitioned into problems with authenticity,
integrity and ordering.  Denial of service may take the form of
discarding information before it reaches its destination or some degree
of delay in delivering information.  Finally, spurious association may
occur when a previous legitimate association initiation is played back
or an initiation is made under false identity.  Security measures may
take the form of either detection or prevention of each of these
threats.  Within the scope of this work, we must identify those threats
that are both of concern and that we expect to be able to mediate.

Of these threats, the passive threats to privacy or confidentiality and
the active threats of authenticity and integrity are probably the most
important to consider here.  To the extent that spurious association
causes threats to the privacy, authenticity, or integrity with respect
to information within servers managing data, it is also important.
Because updates to hint information are idempotent, at least within
short periods of time, we will set aside the problems of ordering for
this analysis.  Denial of service is probably the most difficult of
these areas of threats both to detect and to prevent, and we will
therefore set it aside for the present as well, although it will be seen
that solutions to other problems will also mitigate some of the problems
of denial of service.  Furthermore, because this is intended to be
provide a global service to meet the needs of a variety of communities,
the engineering tradeoffs will be different for different clients.
Hence the rubrics are stated in terms of, "It must be possible..."  It
is important to note that the information of concern here is hint
information, which by nature is not guaranteed to be correct or
up-to-date; therefore, it is unlikely to be worth putting too much
expense into the correctness of hints, because there is no guarantee
that they are still correct anyway.  But the exact choice of degree of
privacy, authenticity, and integrity must be determined by the needs of
the client and the availability of services from the server.

To avoid confusion it is valuable to highlight the meanings of temrs
that have different meanings in other contexts.  In this case, the term
"authoritative" as it is used here connotes the taking of an action or
stamp of approval by a principal (again in the security sense) that has
the right to perform such an act of approval.  It has no implication of
correctness of information, but only perhaps an implication of who

                               - 13 -

claimed it to be correct.  In contrast, the term is often also used
simply to refer to a primary copy of a piece of information for which
there may also be secondary or cached copies available.  In this
discussion of security we use the former meaning, although it may also
be important to be able to learn about whether a piece of information is
from a primary source or not and request that it be primary.

It is also important to distinguish various possible meanings for
"access control."  There are two areas in which distinctions can be
made.  First, there is the question of the kind of access control that
is being addressed, for example, in terms of hints whether it is read
access, read and modify access, or read with verification for
authenticity.  Second, there is the question of to what access is being
controlled.  In the context of naming it might be the names themselves
(not the case for URNs), the mapping of URNs to hints (the business of
an RDS), the mapping of URNs to addresses (not the business of an RDS as
will be discussed below in terms of privacy), or the resource itself
(unrelated to naming or name resolution at all).  We attempt to be clear
about what is meant when using "access control."

There is one further issue to address at this point, the distinction
between mechanism and policy.  In general, a policy is realized by means
of a set of mechanisms.  In the case of an RDS there may be policies
internal to the RDS that it needs to have supported in order to do its
business as it sees fit.  Since, in general it is in the business of
storing and distributing information, most of its security policies may
have to do with maintaining its own integrity, and are rather limited.
Beyond that, to the degree possible, it should impose no policy on its
customers, the publishers and users.  It is they that may have policies
that they would like supported by the RDS.  To that end, an RDS should
provide a spectrum of "tools" or mechanisms that the customers can cause
to be deployed on their behalf to realize policies.  An RDS may not
provide all that is needed by a customer.  A customer may have different
requirements within his or her administrative bounds than outside.
Thus, "it must be possible..."  captures the idea that the RDS must
generally provide the tools to implement policies as needed by the
customers.

The first approach to URN resolution is to discover local hints.  In
order for hints to be discovered locally, they will be as widely
distributed to what is considered to be local for every locale.  The
drawback of such wide distribution is the wide distribution of updates,
causing network traffic problems or delays in delivering updates.  An
alternative model would concentrate hint information in servers, thus
requiring that update information only be distributed to these servers.
In such a model the vulnerable points are the sources of the information
and the distribution network among them.  Attackers on the integrity of
the information stored in a server may come in the form of other a fake
owner of the information or a fake server to the extent that servers
exchange updates with each other.  Wide replication of information among
servers increases the difficult of masquerading at all the locations of
the information as well as reducing the threat of denial service.  These
lead us to three identifiable goals for our security model:



                               - 14 -

* ACCESS CONTROL ON HINTS: It must be possible to create an
  authoritative version of each hint with change control limited only
  to those principals with the right to modify it.  The choice of who
  those principals are or whether they are unlimited must be should by
  the publisher of a hint.

* SERVER AUTHENTICITY: Servers and clients must be able to learn the
  identity of the servers with which they communicate.  This will be a
  matter of degree and it is possible that there will be more
  trustworthy, but less accessible servers, supported by a larger
  cluster of less authenticatable servers that are more widely
  available.  In the worst case, if the client receives what appears to
  be unvalidated information, the client should assume that the hint
  may be inaccurate and confirmation of the data might be sought from
  more reliable but less accessible sources.

* SERVER DISTRIBUTION: Broad availability will provide resistance to
  denial of service.  It is only to the extent that the services are
  available that they provide any degree of trustworthiness.  In
  addition, the distribution of services will reduce vulnerability
  of the whole community, by reducing the trust put in any single
  server.  This must be mitigated by the fact that to the extent trust
  is based on a linked set of servers, if any one fails, the whole
  chain of trust fails; the more elements there are in such a chain,
  the more vulnerable it may become.

Privacy is a more difficult problem to address.  It may be a
double-edged sword; for example, an organization may consider it
critically important that its competitors not be able to read its
traffic, while it may also consider it important to be able to monitor
exactly what its employees are transmitting to and from whom, for a
variety of reasons such as reducing the probability that its employees
are giving or selling the company's secrets to verifying that employees
are not using company resources for private endeavor.  Thus, although
there are likely to be needs for privacy and confidentiality, what they
are, who controls them and how, and by what mechanisms vary widely
enough that it is difficult to say anything concrete about them here.

The privacy of publishers is much easier to safeguard.  Since they are
trying to publish something, in general privacy is probably not desired.
However, publishers do have information that they might like to keep
private: information about who their clients are, and information about
what names exist in their namespace.  The information about who their
clients are may be difficult to collect depending on the implementation
of the resolution system.  For example, if the resolution information
relating to a given publisher is widely replicated, the hits to _each_
replicated copy would need to be recorded.  Of course, determining if a
specific client is requesting a given name can be approached from the
other direction, by watching the client as we saw above.

The other privacy issue for publishers has to do with access control
over URN resolution.  This issue is dependent on the implementation of
the publisher's authoritative (in the sense of "primary) URN resolver
server.  URN resolver servers can be designed to require proof of
identity in order to be issued resolution information; if the client
does not have permission to access the URN requested, the service denies

                               - 15 -

that such a URN exists.  An encrypted protocol can also be used so that
both the request and the response are obscured.  Encryption is possible
in this case because the identity of the final recipient is known (i.e.
the URN server).  Thus, access control over URN resolution can and
should be provided by resolver servers rather than an RDS.

. The Framework

With these assumptions and guidelines in mind, we can conclude with a
general framework within which RDS designs can fall.  As stated earlier,
although this framework is put forth as a suggested guide for RDS
designers, compliance with it will in no way guarantee compliance with
the rubrics.  Such an evaluation must be performed separately.  It is
also understood that there may be RDS services that do not meet the
guidelines in clearly identified ways.  This may be true especially with
early plans and experiments.  For example, although a careful threat
analysis may have been done to understand security requirements, not all
those security issues may be addressed, in order to use existing
facilities to allow for early deployment for experimentation purposes.
All such lack of compliance should be clearly documented.

The design of the framework is based on a simple assumption about the
syntax of a URN a documented in RFC-2141[RFC2141].  This assumed syntax
is:

        URN:<NID>:<NSS>

where URN: is a prefix on all URNs, NID is the namespace identifier, and
NSS is the namespace specific string.  The prefix identifies each URN as
such.  The NID determines the general syntax for all URNs within its
namespace.  The NSS is probably partitioned into a set of delegated and
subdelegated namespaces, and this is probably reflected in further
syntax specifications.  In more complex environments, each delegated
namespace will be permitted to choose the syntax of the variable part of
the namespace that has been delegated to it.  In simpler namespaces, the
syntax will be restricted completely by the parent namespace.  For
example, although the DNS does not meet all the requirements for URNs,
it has a completely restricted syntax, such that any further structuring
must be done only by adding further refinements to the left, maintaining
the high order to low order, right to left structure.  A delegated
syntax might be one in which a host is named by the DNS, but to the
right of that and separated by an "@" is a string whose internal
ordering is defined by the file system on the host, which may be defined
high order to low order, left to right.  Of course, much more complex
and nested syntaxes should be possible, especially given the need to
grandfather namespaces.  In order to resolve URNs, rules will be needed
for two reasons.  One is simply to canonicalize those namespaces that do
not fall into a straightforward (probably right to left or left to
right) ordering of the components of a URN, as determined by the
delegated naming authorities involved.  It is also possible that rules
will be needed in order to derive from URNs the names of RDS servers to
be used in stages.

The NID defines a top level syntax.  This syntax will determine whether
the NID alone or in conjunction with some extraction from the NSS (for
the top level naming authority name) is to be used to identify the first

                               - 16 -

level server to be contacted.  At each stage of the lookup either a new
rule for generating the strings used in yet another lookup (the strings
being the identity of another RDS server and possibly a string to be
resolved if it is different than the original URN) or a reference
outside the RDS to a URN resolver service, sidestepping any further use
of the RDS scheme.  Figure 1 depicts this process.


                            URN:<NID><NSS>
                                 |
                                 |
                                 |
                                 |
                                 v
                       +-------------------+
                       |Global NID registry|
                       +-------------------+
                                 |
                                 |
                                 |
              (return rule or URN resolver service reference)
                                 |
                                 +----------------------------------+
                                 |                                  |
                       +->(apply rule to determine RDS server)      |
                       |         |                                  |
                       |         |                                  |
                       |         |                                  |
                       |    +----------+                            |
                       |    |RDS server|          +-----------------+
                       |    +----------+          |
                       |      |   |               v
                       |      |   |   (set of choices)
                       |      |   +----+----------(...)--------+
                       |   (rule)      |                       |
                       |      |        |                       |
                       |      |        |                       |
                       +------+        |                       |
                                       v                       v
                                  +----------+            +----------+
                                  |URN       |            |URN       |
                                  |resolver  |            |resolver  |
                                  |service   |            |service   |
                                  +----------+            +----------+



        Figure 1: An RDS framework


There are several points worth noting about the RDS framework.  First,
it leaves open the determination of the protocols, data organization,
distribution and replication needed to support a particular RDS scheme.
Second, it leaves open the location of the computations engendered by
the rules.  Third, it leaves open the possibility that partitioning
(distribution) of the RDS database need not be on the same boundaries as

                               - 17 -

the name delegation.  This may seem radical to some, but if the
information is stored in balanced B-trees for example, the partitioning
may not be along those naming authority delegation boundaries (see
[Sl97]).  Lastly, it leaves open access to the Global NID Registry.  Is
this distributed to every client, or managed in widely distributed
servers?

One concept that has not been addressed in Figure 1 is that there may be
more than one RDS available at any given time, in order to allow for
evolution to new schemes.  Thus, the picture should probably look more
like Figure 2.


                         URN:<NID>:<NSS>
                               |
                               |
                   +-----------+-------(...)-------+
                   |                               |
                   |                               |
                   |                               |
                   v                               v
         +---------------------+        +---------------------+
         |Global NID registry 1|        |Global NID registry N|
         +---------------------+        +---------------------+
                   .                               .
                   .                               .
                   .                               .


        Figure 2: More than one co-existing RDS scheme


If we are to support more than one co-existing RDS scheme, there will
need to be coordination between them with respect to storage and
propagation of information and modifications.  The issue is that
generally it should be assumed that all information should be available
through any operational RDS scheme.  One cannot expect potential
publishers to submit updates to more than one RDS scheme.  Hence there
will need to be a straightforward mapping of information from one to the
other of these schemes.  It is possible that that transformation will
only go in one direction, because a newer RDS service is replacing an
older one, which is not kept up to date, in order to encourage transfer
to the newer one.  Thus, at some point, updates may be made only to the
newer one and not be made available to the older one, as is often done
with library catalogs.

This framework is presented in order to suggest to RDS scheme designers
a direction in which to start designing.  It should be obvious to the
reader that adherence to this framework will in no way guarantee
compliance with the guidelines or even the assumptions described in
Sections 2 and 3.  These must be reviewed independently as part of the
design process.  There is no single correct design that will conform to
these guidelines.  Furthermore, it is assumed that preliminary proposals
may not meet all the guidelines, but should be expected to itemized and
justify any lack of compliance.

                               - 18 -

. Acknowledgments

Foremost acknowledgment for this document goes to Lewis Girod, as my
co-author on a preliminary URN requirements document and for his
insightful comments on this version of the document.  In addition, I
recognize the contributors to a previous URN framework document, the
"Knoxville" group.  There are too many of you to acknowledge here
individually, but thank you.  Finally, I must thank the contributors to
the URN working group mailing list (urn-ietf@bunyip.com), for their
animated discussions on these and related topics.

. References

[RFC1736] Kunze, J., "Functional Recommendations for Internet Resource
Locators", RFC 1736, February, 1995.

[RFC1737] Sollins, K. and Masinter, L., "Functional Requirements for
Uniform Resource Names", RFC 1738, December, 1994.

[RFC1738] Berners-Lee, T., Masinter, L., McCahill, M., "Uniform
Resource Locators (URL)", RFC 1738, December, 1994.

[RFC2141] Moats, Ryan, "URN Syntax", RFC 2141, May 1997.

[Sl97] Slottow, E.G., "Engineering a Global Resolution Service,"
MIT-LCS-TR712, June, 1997.  Currently available as
<http://ana.lcs.mit.edu/anaweb/ps-papers/tr-712.ps> or
<http://ana.lcs.mit.edu/anaweb/pdf-papers/tr712.pdf>.

[VK83] Voydock, V. L., and Kent, S. T., "Security Mechanisms in
High-Level Protocols", ACM Computing Surveys, v. 15, No. 2, June,
1983, pp. 135-171.

. Contact information:

Karen Sollins
MIT Laboratory for Computer Science
 Technology Sq.
Cambridge, MA 02139

Tel: +1 617 253 6006
Email: sollins@lcs.mit.edu

This Internet Draft expires on December 4, 1997.













                               - 19 -