wdiff draft-ietf-ecrit-mapping-arch-00.txt draft-ietf-ecrit-mapping-arch-01.txt

ECRIT H. Schulzrinne
Internet-Draft Columbia U.
Intended status: Informational December 17, 2006
Expires: February 8, June 20, 2007 August 7, 2006

Location-to-URL Mapping Architecture and Framework
draft-ietf-ecrit-mapping-arch-00
draft-ietf-ecrit-mapping-arch-01

Status of this Memo

By submitting this Internet-Draft, each author represents that any
applicable patent or other IPR claims of which he or she is aware
have been or will be disclosed, and any of which he or she becomes
aware will be disclosed, in accordance with Section 6 of BCP 79.

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/ietf/1id-abstracts.txt.

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

This Internet-Draft will expire on February 8, June 20, 2007.

Abstract

This document describes an architecture for a global, scalable,
resilient and administratively distributed system for mapping
geographic location information to URLs. URLs, using the Location-to-
Service (LoST) protocol. The architecture generalizes well-known
approaches found in hierarchical lookup systems such as DNS. The architecture does not depend on using a
specific protocol, but does require that protocols can summarize the
coverage region of a node.

Table of Contents

1. The Mapping Problem . . . . . . . . . . . . . . . . . . . . . 3
2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . 3 . 4
3. Definitions . . . . . . . . . . . . . . . . . . . . . . . . 3
4. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 5
4.1 4
4. Overview of Operation . . . . . . . . Architecture . . . . . . . . . . 5
4.2 Seekers: The Users of the Mapping System . . . . . . . . . 5
4.3 Trees: Authoritative Knowledge . . . . . . . . . . . . . . 6
4.4 Forest Guides: Finding the Right Tree . . . . . . . . . . 7
4.5 Resolvers: Finding Forest Guides and Caching Data . . . . 7
4.6
4.1. Minimal System Architecture . . . . . . . . . . . . . . . 7 6
5. Seeker . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 . 6
6. Resolver . . . . . . . . . . . . . . . . . . . . . . . . . . 8
7. Trees . . . . . . . . . . . . . . . . . 7
7. Trees: Maintaining Authoritative Knowledge . . . . . . . . . . 8
7.1
7.1. Basic Operation . . . . . . . . . . . . . . . . . . . . . 8
7.2
7.2. Answering Queries . . . . . . . . . . . . . . . . . . . . 10
7.3
7.3. Overlapping Coverage Regions . . . . . . . . . . . . . . . 11
7.4 10
7.4. Scaling and Reliability . . . . . . . . . . . . . . . . . 11
8. Forest Guides . . . . . . . . . . . . . . . . . . . . . . . . 11
9. Security Considerations . Configuring Service Numbers . . . . . . . . . . . . . . . . . 12
10. IANA Security Considerations . . . . . . . . . . . . . . . . . . . . 13 14
11. References IANA Considerations . . . . . . . . . . . . . . . . . . . . . 15
12. Acknowledgments . . . . 13
11.1 Normative References . . . . . . . . . . . . . . . . . . 13
11.2 Informative . 15
13. References . . . . . . . . . . . . . . . . . 13
Author's Address . . . . . . . . . 15
13.1. Normative References . . . . . . . . . . . . . . . 14
A. Configuring Emergency Dial Strings . . . . 15
13.2. Informative References . . . . . . . . . . . 14
B. Acknowledgments . . . . . . . 15
Author's Address . . . . . . . . . . . . . . . . . . . . . . . . . 16
Intellectual Property and Copyright Statements . . . . . . . . . . 17

1. The Mapping Problem

One of the central problems of providing emergency services to
Internet systems

It is often desirable to map geographic location allow users to access a set of emergency
services, represented by PSAPs, service that can provide assistance for
provides a common function, but is actually offered by a variety of
local service providers. In many of these cases, the service
provider chosen depends on the location of the person wishing to
access that
particular service. Among the best-known public services of this
kind is emergency calling, where emergency calls are routed to the
most appropriate public safety answering point (PSAP), based on the
caller's physical location. Other services, from food delivery to
directory services and roadside assistance, also follow this general
pattern. This is a mapping problem, problem [8], where a geographic location
and a service identifier (URN) [10] is translated into a set of URIs URIs,
the service URIs, that allow the Internet system to contact an
appropriate network entity. Other services may
also find such location-to-URI mappings entity that provides the service.

The caller does not need to know where the service is being provided
from, and the location of use. the service provider may change over time,
e.g., to deal with temporary overloads, failures in the primary
service provider location or long-term changes in system
architecture. For emergency services, this problem is described in
more detail in [6].

The overall emergency calling architecture [6] separates mapping from
placing calls or otherwise invoking the service, so the same
mechanism can be used to verify that a mapping exists ("address
validation") or to obtain test service URIs.

Mapping locations to URIs describing services requires a distributed,
scalable and highly resilient infrastructure. Authoritative
knowledge about such mappings is distributed among a large number of
autonomous entities that may have no direct knowledge of each other.
In this document, we describe an architecture for such a global
service. It allows significant freedom to combine and split
functionality among actual servers and imposes few requirements as to
who should operate particular services.

Besides determining the PSAP service URI, end systems also need to
determine the local emergency dial strings. service numbers. As discussed in Appendix A, Section 9, the
architecture described here can also address that problem.

The architecture described below does not depend on a particular
mapping protocol, but naturally assumes that such protocols provide
certain features, such as the ability to discover here uses the coverage region Location-to-Service
Translation (LoST) [2] protocol, although much of tree nodes. In this introduction, we describe the four
participants in discussion
would also apply for other mapping protocols satisfying the system at a high level. Each role will later be
introduced in more detail. mapping
requirements [8].

2. Terminology

In this document, the key words "MUST", "MUSTNOT", "REQUIRED",
"SHALL", "SHALLNOT", "SHOULD", "SHOULDNOT", "RECOMMENDED", "MAY", and
"OPTIONAL" are to be interpreted as described in RFC 2119 [1] and
indicate requirement levels for compliant implementations.

3. Definitions

[Note: The terminology below is still evolving and needs
refinement.]

In addition to the terms defined in [11], [8], this document uses the
following terms to describe LoST: LoST clients and servers:

authoritative mapping server (AMS): Resolver An authoritative mapping server
(AMS) is a LoST server that can provide the authoritative answer
to a particular set of queries, e.g., covering a set of PIDF-LO
civic labels or a particular region described by a geometric
shape. In some (rare) cases of territorial disputes, two
resolvers may be authoritative for the same region. An AMS may
redirect or forward a query to other AMS within the tree.
caching resolver: A caching resolver is contacted by a seeker,
consults a forest mapping server and then resolves the query using
an appropriate tree.
child: A child is a resolver an AMS that is authoritative for a subregion of
a particular server.
another AMS. A child can in turn be parent. parent for another AMS.
(tree node) cluster: A node cluster is a group of resolver (servers) LoST servers that
all share the same mapping information and return the same results
for queries. Clusters provide redundancy and share query load.
Clusters are fully-meshed, i.e., they all exchange updates with
each other.
complete: A civic mapping region is considered complete if it covers
a set of hierarchical labels in its entirety, i.e., there is no
other resolver that covers parts of the same region. (A complete
mapping may have children that cover strict subsets of this
region.) For example, a region spanning the whole country is
complete, but a region spanning only some of the streets in a city
is not.
forest guide: A forest guide has knowledge of the coverage region of
all trees.
mapping: A mapping is a short-hand for 'mapping from a location
object to one or more URLs describing either another mapping
server or the desired PSAP URLs.' service URLs'.
parent: A mapping server that covers the region of all of its
children. A mapping server without a parent is a root resolver.
peer:
resolver: A resolver maintains associations other resolvers, called
peers. Peers synchronize their region maps. is contacted by a seeker, consults a forest
mapping server and then resolves the query using an appropriate
tree.
seeker: The resolver, ESRP or end system A seeker is a LoST client requesting a mapping. A seeker
does not provide mapping services to others, but may cache results
for its own use.
region map: A data object describing a contiguous area covered by a
resolver, an
AMS, either as a subset of a civic address or a geometric object.
root region map: A data object describing a contiguous area covered
by a resolver, with no parent map.
resolver: The server providing (part of) the mapping service. Resolvers cooperate to offer the contact authoritative mapping service servers to seekers. answer
queries by seekers, and may cache query results.

tree: A tree consists of a self-contained hierarchy of authoritative
mapping servers. Each tree exports its coverage region to the
forest mapping servers.

4. Introduction

4.1 Overview of Operation Architecture

In short, end users of the LoST-based [2] mapping mechanism, called
seekers, contact resolvers that cache query results and know one or
more "forest guides". Forest guides know the coverage region of
trees and direct queries to the node at the top of the appropriate
tree. Trees consist of authoritative mapping servers and maintain
the authoritative mapping information. Figure 1 shows the
interaction of the components.

/-\ /-\ +-----+ +-----+
| S +******* R ********* FG *-----------------+ FG |
\-/ \-/ | |* | |
+--+--+ * +--+--+
| * |
| * |
| * |
| * |
/-\ +--+--+ * +--+--+
| R +------>+ FG +-----*-----------+ FG |
\-/ | | * | |
+--+--+ * +--+--+
| * |
| * |
| * |
|*** ^
/ \ / \
/ \ / \
/ \ / \
/ \ / \
----------- -----------
tree tree

Architecture diagram, showing seekers (S), resolvers (R), forest
guides (FG) and trees. The star (*) line indicates the flow of the
query and responses in recursive mode.

Figure 1

4.2 Seekers: The Users of the Mapping System

Clients desiring mappings are known as seekers. Thus, seekers are
the end users of the mapping information. Examples of such clients
include SIP proxy servers or SIP end systems wishing to place an
emergency call. Seekers provide location information describing a
small geographic area and obtain one or more URIs describing the
service. Seekers may need to obtain this information in several
steps, i.e., they may obtain pointers to intermediate servers that
lead them closer to the final mapping. Seekers MAY cache query
results for later use, but otherwise have no obligations to other
entities in the system.

4.3 Trees: Authoritative Knowledge

The architecture assumes that authoritative knowledge about the
mapping data is distributed among many independent administrative
entities, but clients (seekers) needing the information may
potentially need to find out mapping about any spot on earth.
(Extensions to extra-terrestrial applications are left for future
exploration.) Different types of services may divide responsibility
differently and are independent of each other. Each node
participating in the system has authoritative knowledge about
mappings within its coverage region, typically, but not necessarily,
a contiguous geographic region described by a polygon in geospatial
coordinates or a set of civic address descriptors (e.g., "country =
DE, A1 = Bavaria"). These coverage regions may be aligned with
political boundaries, but that is not required. In most cases, to
avoid confusion, only one node is responsible for a particular
geographic or civic location, but the system can also deal with cases
where coverage regions overlap.

The architecture assumes that knowledge about mappings is
hierarchical, represented as a tree. Each tree node knows the
coverage region of its children and sends queries to the appropriate
server "down" the tree. There are no assumptions about the coverage
region of a tree. For example, a tree could cover a single city, or
a state/province or a whole country. Nodes within a tree need to
loosely coordinate their operation, but they do not need to be
operated by the same administrator.

Thus, the mapping function for the world is divided among trees. The
collection of trees may not cover the whole world and trees are added
and removed as the organization of mapping data changes. We call the
collection of trees a forest. There is no limit on the number of
trees within the forest, but the author pictures guesses that the number of
trees will likely be somewhere between a few hundred and a few
thousand. The lower estimate would apply if each country operates
one tree. We assume that tree coverage information changes
relatively slowly, on the order of a few changes less than one change per year per
tree, although the system imposes no specific threshold. (To be sure,
information within a tree is likely to change much more frequently.)

4.4 Forest Guides: Finding the Right Tree

Unfortunately, just having trees covering various regions of the
world is not sufficient as a client of the mapping protocol
coverage would not
generally be able to keep track of all the trees in the forest. To
facilitate orientation among the trees, we introduce change, for example, if a "forest
guide". It country is a server that keeps track of the coverage split or merged
or if two trees for different regions become part of
the trees. For scalability and reliability, there will need to be a
large number of forest guides, all providing the same information. A
seeker can contact any forest guide and will then be directed to the
right tree or, rarely, set of trees.

4.5 Resolvers: Finding Forest Guides and Caching Data

A seeker can contact a forest guide directly, but may not be able to
easily locate such a guide. In addition, seekers in the same
geographic area may already have asked larger tree.
(On the same question. Thus, it
makes sense to introduce another entity, other hand, information within a resolver, that knows how tree is likely to contact one or change
much more forest guides and caches earlier queries to
accelerate the response to mapping queries.

4.6 frequently.)

4.1. Minimal System Architecture

It is possible to build a functioning system consisting only of
seekers and resolvers if these resolvers have other means of
obtaining mapping data. For example, a company acting as a mapping
service provider could collect mapping records manually and make them
available to their customers through the resolver. While feasible as
a starting point, such an architecture is unlikely to scale globally.
Among other problems, it becomes very hard for providers of
authoritative data to ensure that all such providers have up-to-date
information. If new trees are set up, they would somehow make
themselves known to these providers. Such a mechanism would be
similar to the old "hosts.txt" mechanism for distributing host
information in the early Internet. Internet before DNS was developed.

Below, we describe the operation of each component in more detail.

5. Seeker

Clients desiring location-to-service mappings are known as seekers.
Seekers are consumers of mapping data and originate queries. LoST queries as
LoST protocol clients. Seekers do not answer LoST queries. They
contact either forest guides or resolvers to find the appropriate
tree that can authoritatively answer their questions. As noted in the introduction, seekers Seekers can be
end systems or call routing entities such as SIP proxy servers.

Seekers may need to obtain mapping information in several steps,
i.e., they may obtain pointers to intermediate servers that lead them
closer to the final mapping. Seekers MAY cache query results for
later use, but otherwise have no obligations to other entities in the
system.

Seekers need to be able to identify appropriate resolvers. The
mechanism for providing seekers with that information is likely to
differ depending on who operates the resolvers. For example, if the
voice service provider operates the resolver, it might include the
location of the resolver in the SIP configuration information it
distributes to its user agents. An Internet access provider might or
enterprise can provide a pointer to a resolver via DHCP. DHCP [5]. In an
ad-hoc or zero-
configuration zero-configuration environment, appropriate service
directories may advertise resolvers.

For emergency calling, seekers could issue queries at boot time,
periodically when cached information expires or only when placing an
emergency call. It is probably unnecessary to continuously update
mapping information for seekers representing a small user population,
e.g., a single phone or residential SIP proxy.

Like other entities in the system, seekers can cache responses. This
is particularly useful if the response describes the result for a
civic or geospatial region, not rather than just a point. For example,
for mobile nodes, seekers would only have to update their resolution
results when they leave the coverage area of a service provider, such
as a PSAP for emergency services, and can avoid repeatedly polling
for this
information. information whenever the location information changes
slightly. (Mobile nodes would also need a location update mechanism
that is either local or triggered when they leave the current service
area.) This will likely be of particular benefit for seekers
representing a large user population, such as the outbound proxy in a
corporate network. For example, rather than having to query
separately for each cubicle, information provided by the
authoritative node may indicate that the whole campus is covered by
the same PSAP.

6. Resolver

Resolvers mediate between seekers service provider.

Given this caching mechanism and forest guides. Their primary
role is cache lifetimes of several days,
most mobile users traveling to avoid having seekers find forest guides on and from work would only need to
obtain service area information along their own.
Unlike commute route once during
each cache lifetime.

6. Resolver

A seeker can contact a forest guides, resolvers do not store worldwide coverage maps, guide (see below) directly, but they may cache regions returned as part of query results.

As noted earlier, not
be able to easily locate such a guide. In addition, seekers can in the
same geographic area may already have asked the same question. Thus,
it makes sense to introduce another entity, known as a resolver in
the architecture, that knows how to contact one or more forest guides directly.
and caches earlier queries to accelerate the response to mapping
queries and to improve the resiliency of the system.

From a protocol perspective, a resolver acts in the same way as a
seeker, except that it knows one or more forest guide. guides.

ISPs or VSPs would include the address of a suitable resolver in
their configuration information, i.e., in SIP configuration for a VSP
or DHCP [5] for an ISP. Resolvers are manually configured with the
name of one or more forest guides.

7. Trees

7.1 Trees: Maintaining Authoritative Knowledge

7.1. Basic Operation

As noted in

The architecture assumes that authoritative knowledge about the introduction, trees are
mapping data is distributed among many independent administrative
entities, but clients (seekers) needing the authoritative source of information may
potentially need to find out mapping data. about any spot on earth.
(Extensions to extra-terrestrial applications are left for future
exploration.) Information is organized hierarchically, in a tree,
with tree nodes representing larger geographic areas pointing to
several child nodes each representing a smaller area. Each tree node
can be a cluster of LoST servers that all contain the same
information and back up each other.

Each tree can map a location described by civic and geographic
coordinates for one type of service (such as 'police' 'sos.police', 'sos.fire'
or
'fire'), 'counseling'), although nothing prevents re-using the same tree
for multiple different services. The collection of all trees for one
service is known as a forest.

Each tree node cluster knows the coverage region of its children and
sends queries to the appropriate server "down" the tree. Each such
tree node knows authoritatively about the service mappings for a
particular region, typically, but not necessarily, contiguous. The
region can be described by a polygon in geospatial coordinates or a
set of civic address descriptors (e.g., "country = DE, A1 =
Bavaria"). These coverage regions may be aligned with political
boundaries, but that is not required. In most cases, to avoid
confusion, only one cluster is responsible for a particular
geographic or civic location, but the system can also deal with cases
where coverage regions overlap.

There are no assumptions about the coverage region of a tree as a
whole. For example, a tree could cover a single city, or a state/
province or a whole country. Nodes within a tree need to loosely
coordinate their operation, but they do not need to be operated by
the same administrator.

The tree architecture is roughly similar to the domain name system
(DNS), except that delegation is not by label, but rather by region.
(Naturally, DNS does not have the notion of forest guides.) One can
also draw analogies to LDAP, when deployed in a distributed fashion.

Tree nodes maintain two types of information, namely coverage regions
and mappings. Coverage regions describe the region served by a child
node in the tree and point to a child node for further resolution.
Mappings contain an actual service URI leading to a PSAP service provider
or another signaling server representing a group of PSAPs. To provide
redundancy, a mapping entry can also contain multiple URLs,
indicating both primary and backup services. For example, it service
providers, which in turn might
contain both a local PSAP and a state agency that takes over if the
local PSAP fails. Unlike DNS NAPTR and SRV facilities, these can
survive DNS failures and thus provide an additional, complementary
mechanism further route signaling requests to introduce redundancy services.
more servers covering smaller regions.

Leaf nodes, i.e., nodes without children, only maintain mappings,
while tree nodes above the leaf nodes only maintain coverage regions.
An example for emergency services of a leaf node entry is shown
below, indicating how queries for three towns are directed to
different PSAPs. Queries for Englewood are directed to another LoST
server instead.

country A1 A2 A3 resource
US NJ Bergen Leonia sip:psap@leonianj.gov
US NJ Bergen Fort Lee sip:emergency@fortleenj.org
US NJ Bergen Teaneck sip:police@teanecknjgov.org
US NJ Bergen Englewood lost:englewoodnj.gov
....

Coverage regions are described by sets of polygons enclosing
contiguous geographic areas or by descriptors enumerating groups of
civic locations. For the former, the LoST server performs a point-
in-polygon operation to find the polygon that contains the query
point. (More complicated geometric matching algorithms may be added
in the future.)

For example, a state-level tree node for New Jersey in the United
States may contain the following coverage region entries, indicating
that any query matching a location in Bergen County, for example,
would be redirected or forwarded to the node located at
bergen.nj.example.org. There is no requirement that all child nodes
cover the same level within the civic hierarchy. As an example, in
the table below, the city of Newark has decided to be listed directly
within the state node, rather than through the county. Longest-match
rules allow partial coverage, so that for queries for all other towns
within Essex county would be directed to the county node for further
resolution. In the example below, we use a fictitious URL scheme,
'rp', to identify the resolution protocol. In actual use, each entry
would have one or more URLs pointing to resolution servers for
different protocols. Each entry may also contain multiple URLs for
the same protocol to indicate primary and backup services.

C A1 A2 A3 resource
US NJ Atlantic * rp://atlantic.nj.example.org/sos lost:atlantic.nj.example.org/sos
US NJ Bergen * rp://bergen.nj.example.org/sos lost:bergen.nj.example.org/sos
US NJ Monmouth * rp://monmouth.nj.example.org/sos lost:monmouth.nj.example.org/sos
US NJ Essex * rp://essex.nj.example.org/sos lost:essex.nj.example.org/sos
US NJ Essex Newark rp//newark.example.com/sos lost:newark.example.com/sos
....

Thus, there is no substantial difference between coverage region and
mapping data. The only difference is that coverage regions return
mapping protocol
LoST URLs, while mapping entries contain PSAP service URLs. Mapping
entries may be specific down to the house or floor level or may only
contain street-level information. For example, in the United States,
civic mapping data for emergency services is generally limited to
address ranges ("MSAG data"), so initial mapping databases may only
contain street-level information.

To automate operations, a suitable mapping protocol would thus need
to be able the maintenance of trees, the LoST synchronization
mechanism [11] allows nodes to query other nodes for their mapping data and
coverage region. regions. In the example above, the state-run node would
query the county nodes and thus
aggregate use the coverage data. records returned to distribute
incoming LoST queries to the county nodes. Conversely, nodes could
also contact their parent nodes. nodes to tell them about their coverage
region. There is some benefit of child nodes contacting their
parents, as this allows changes in coverage region to propagate
quickly up the tree.

7.2

7.2. Answering Queries

Within a tree, the basic operation is straightforward: A query
reaches the root of the tree. That node determines which coverage
region matches that request and forwards the request to the URL
indicated in the coverage region record, returning a response to the
querier when it in turns receives an answer (recursion).
Alternatively, the node returns the URL of that child node to the
querier.
querier (iteration). This process applies to each node, i.e., a node
does not need to know whether the original query came from a parent
node, a seeker, a forest guide or a resolver.

For efficiency, a node MAY return region information instead of a
point answer. Thus, instead of returning that a particular
geospatial coordinate maps to a service or mapping URL, it MAY return
a polygon indicating the region for which this answer would be
returned, along with expiration time (time-to-live) information. The
querying node can then cache this information for future use.

For civic coordinates, trees may not include individual entries mapping
records for each floor, house number or street. To avoid giving the
wrong indication that a particular location has been found valid, the
protocol SHOULD return an indication
LoST can indicate which parts of the location information have
actually been mapped.

7.3 used to look up a mapping.

7.3. Overlapping Coverage Regions

In some cases, coverage regions may overlap, either because there is
a dispute as to who handles a particular geographic region or, more
likely, since the resolution of the coverage map may not be
sufficiently high. For example, a node may "shave some corners" off
its polygon, so that its coverage region appears to overlap with its
geographic neighbor. For civic coordinates, houses on the same
street may be served by different PSAPs. The mapping mechanism needs
to work even if a coverage map is imprecise or if there are disputes
about coverage.

The solution for overlapping coverage regions is relatively simple.
If a query matches multiple coverage regions, the node returns all
URLs, in redirection mode, or queries both children, if in recursive
mode. If the overlapping coverage is caused by imprecise coverage
maps, only one will return a result and the others will return an
error indication. If the particular location is disputed territory,
the response will contain all answers, leaving it to the querier to
choose the preferred solution or trying to contact all services in
turn.

7.4

7.4. Scaling and Reliability

Since they provide authoritative information, tree nodes need to be
highly reliable. Thus, while this document refers to tree nodes as
logical entities within the tree, an actual implementation would
likely replicate node information across several servers, forming a
cluster. Each such node would have the same information. Standard
techniques such as DNS SRV records can be used to select one of the
servers. Replication within the cluster can use any suitable
protocol mechanism, but a standardized incremental update mechanism
makes it easier to spread those nodes across multiple independently-
administered locations. The techniques developed for meshed SLP [7] [3]
are applicable here.

8. Forest Guides

Unfortunately, just having trees covering various regions of the
world is not sufficient as a client of the mapping protocol would not
generally be able to keep track of all the trees in the forest. To
facilitate orientation among the trees, we introduce a "forest
guide". It is a server that keeps track of the coverage regions of
the trees. For scalability and reliability, there will need to be a
large number of forest guides, all providing the same information. A
seeker can contact any forest guide and will then be directed to the
right tree or, rarely, set of trees. Forest guides distribute records describing do not provide
mapping information themselves.

Introducing forest guides avoids creating a global root, with the coverage region
attendant management and control issues. Trees can also restrict
their cooperation to parts of the information. For example, if
country C does not recognize country T, C can propagate tree regions
for
trees. all but T.

For authenticity, the records are SHOULD be digitally signed. They are
used by resolvers and possibly seekers to find the appropriate tree
for a particular area. All forest guides should have consistent
information, i.e., a collection of all the coverage regions of all
the trees. A tree node at the top of a tree can contact any forest
guide and inject new coverage region information into the system.
One would expect that each tree announces its coverage to more than
one forest guide. Each forest guide peers with one or more other
guides and distributes new coverage region announcements to all other
guides.

Forest guides fulfill a similar role to root servers in DNS.
However, their number is likely to be larger, possibly counted in
hundreds. They distribute information, signed for authenticity,
offered by trees.

Forest guides can, in principle, be operated by anybody, including
voice service providers, Internet access providers, dedicated
services providers and enterprises.

As in routing, peering with other forest guides implies a certain
amount of trust in the peer. Thus, peering is likely to require some
negotiation between the administering parties concerned, rather than
automatic configuration. The mechanism itself does not imply a
particular policy as to who gets to advertise a particular coverage
region.

9. Configuring Service Numbers

The section below is not directly related to the problem of
determining service location, but is an instance of the more generic
problem solved by this architecture, namely mapping location
information to service-related parameters, such as service numbers.

For the foreseeable future, some user devices and software will
emulate the user interface of a telephone, i.e., the only way to
enter call address information is via a 12-button keypad with digits
and the asterisk and hash symbol. These devices use service numbers
to identify services. The best-known examples of service numbers are
emergency numbers, such as 9-1-1 in North America and 1-1-2 in
Europe. However, many other public and private service numbers have
been defined, ranging in the United States from 3-1-1 for non-
emergency local government services to 4-1-1 for directory assistance
to various "800" numbers for anything from roadside assistance to
legal services to home-delivery food.

Such service numbers are likely to be used until essentially all
communication devices feature IP connectivity and an alphanumeric
keyboard. Unfortunately, for emergency services, more than 60
emergency numbers are in use throughout the world, with many of those
numbers serving non-emergency purposes elsewhere, e.g., identifying
repair or directory services. Countries also occasionally change
their emergency numbers to conform to regional agreements. An
example is the introduction of "1-1-2" for countries in Europe.

Thus, a system that allows devices to be used internationally to
place calls needs to allow devices to discover service numbers
automatically. In the Internet-based system proposed here [6], these
numbers are strictly used as a human user interface mechanism and are
generally not visible in call signaling messages, which carry the
service URN [10] instead.

For the best user experience, systems should be able to discover two
sets of service numbers, namely those used in the user's home country
and in the country the user is currently visiting. The user is most
likely to remember the former, but a companion borrowing a device in
an emergency, say, may only know the local emergency numbers.

Determining home and local service numbers is a configuration
problem, but unfortunately, existing configuration mechanisms are
ill-suited for this purpose. For example, a DHCP server might be
able to provide the local service numbers, but not the home numbers.
When virtual private networks (VPNs) are used, even DHCP may provide
numbers of uncertain origin, as a user may contact to the home
network or some local branch office of the corporate network.
Similarly, SIP configuration [4] would be able to provide the numbers
valid at the location of the SIP service provider, but even a SIP
service provider with national footprint may serve customers that are
visiting any number of other countries.

Also, while initially there are likely to be only a few service
numbers, e.g., for emergency services, the LoST architecture can be
used to support other services, as noted. Configuring every local
DHCP or SIP configuration server with that information is likely to
be error-prone and tedious.

For these reasons, the LoST-based mapping architecture supports
providing service numbers to end systems based on caller location.
The mapping operation is almost exactly the same as for determining
the service URL. The mapping can be obtained either along with
determining the service URL or separately. The major difference
between the two requests is that service numbers often have much
larger regions of validity than the service URL itself. Also, the
service number is likely to be valid longer than the service URL.
Finally, an end system may want to look up the service number for its
home location, not just the current (visited) location.

10. Security Considerations

Security considerations for emergency services mapping are discussed
in [9], while [10] discusses issues related to the service URN, one
of the inputs into the mapping protocol. LoST-related security
considerations are naturally discussed in the LoST [2] specification.

The architecture addresses the following security issues, usually
through the underlying transport security associations:

Server impersonation: Seekers, resolvers, fellow tree guides and
cluster members and peers can assure themselves of the identity of the
remote party by using the facilities in the underlying channel
security mechanism, such as TLS.
Query or query result corruption: To avoid that an attacker can
modify the query or its result, the architecture RECOMMENDS the
use of channel security, such as TLS. Results SHOULD also be
digitally signed, e.g., using XML digital signatures. Note,
however, that simple origin assertion may not provide the end
system with enough useful information as it has no good way of
knowing that a particular signer is authorized to represent a
particular geographic area. It might be necessary that certain
well-known Certificate Authorities (CAs) vet sources of mapping
information and provide special certificates for that purpose. In
many cases, a seeker will have to trust its local resolver to vet
information for trustworthiness; in turn, the resolver may rely on
trusted forest guides to steer it to the correct information.
Region corruption: To avoid that a third party or an untrustworthy
member of a server population introduces a region map that it is
not authorized for, any node introducing a new region map MUST
sign the object by encapsulating the data into a CMS wrapper. A
recipient MUST verify, through a local policy mechanism, that the
signing entity is indeed authorized to speak for that region.
Determining who can speak for a particular region is inherently
difficult unless there is a small set of authorizing entities that
participants in the mapping architecture can trust. Receiving
systems should be particularly suspicious if an existing region
map is replaced with a new one with a new mapping address. In
many cases, trust will be mediated: A seeker will have a trust
relationship with a resolver. The resolver, in turn, will contact
a trusted forest guide.

Additional threats that need to be addressed by operational measures
include denial-of-service attacks.

10. attacks [7].

11. IANA Considerations

Since this document describes an architecture, not a protocol, it
does not ask IANA to register any protocol constants.

11.

12. Acknowledgments

Richard Barnes, Jong Yul Kim, Andrew Newton, Murugaraj Shanmugam,
Richard Stastny, and Hannes Tschofenig provided helpful comments.

13. References

11.1

13.1. Normative References

[1] Bradner, S., "Key words for use in RFCs to Indicate Requirement
Levels", BCP 14, RFC 2119, March 1997.

[2] Droms, R., "Dynamic Host Configuration Hardie, T., "LoST: A Location-to-Service Translation Protocol", RFC 2131,
March 1997.

[3] Gulbrandsen, A., Vixie, P., and L. Esibov, "A DNS RR for
specifying the location of services (DNS SRV)", RFC 2782,
February 2000.

[4] Peterson, J., "A Presence-based GEOPRIV Location Object Format",
RFC 4119, December 2005.

[5] Rosen, B., "Dialstring parameter for the Session Initiation
Protocol Uniform Resource Identifier",
draft-rosen-iptel-dialstring-04
draft-ietf-ecrit-lost-02 (work in progress), June October 2006.

11.2

13.2. Informative References

[6] Guttman, E., Perkins, C., Veizades, J., and M. Day, "Service
Location Protocol, Version 2", RFC 2608, June 1999.

[7]

[3] Zhao, W., Schulzrinne, H., and E. Guttman, "Mesh-enhanced
Service Location Protocol (mSLP)", RFC 3528, April 2003.

[8] Newton, A. and M. Sanz, "IRIS: The Internet Registry
Information Service (IRIS) Core Protocol", RFC 3981,
January 2005.

[9] Krochmal, M. and S. Cheshire, "DNS-Based Service Discovery",
draft-cheshire-dnsext-dns-sd-03 (work in progress), July 2005.

[10]

[4] Petrie, D., "A Framework for Session Initiation Protocol User
Agent Profile Delivery", draft-ietf-sipping-config-framework-08 draft-ietf-sipping-config-framework-09
(work in progress), March October 2006.

[11]

[5] Schulzrinne, H., "A Dynamic Host Configuration Protocol (DHCP)
based Location-to-Service Translation Protocol (LoST)
Discovery Procedure", draft-ietf-ecrit-dhc-lost-discovery-00
(work in progress), December 2006.

[6] Rosen, B., "Framework for Emergency Calling in Internet
Multimedia", draft-ietf-ecrit-framework-00 (work in progress),
October 2006.

[7] Rosen, B. and J. Polk, "Best Current Practice for
Communications Services in support of Emergency Calling",
draft-ietf-ecrit-phonebcp-00 (work in progress), October 2006.

[8] Schulzrinne, H. and R. Marshall, "Requirements for Emergency
Context Resolution with Internet Technologies",
draft-ietf-ecrit-requirements-10
draft-ietf-ecrit-requirements-12 (work in progress), June
August 2006.

[9] Taylor, T., "Security Threats and Requirements for Emergency
Call Marking and Mapping", draft-ietf-ecrit-security-threats-03
(work in progress), July 2006.

[10] Schulzrinne, H., "A Uniform Resource Name (URN) for Services",
draft-ietf-ecrit-service-urn-05 (work in progress),
August 2006.

[11] Schulzrinne, H., "Synchronizing Location-to-Service Translation
(LoST) Servers", draft-schulzrinne-ecrit-lost-sync-00 (work in
progress), November 2006.

Author's Address

Henning Schulzrinne
Columbia University
Department of Computer Science
450 Computer Science Building
New York, NY 10027
US

Phone: +1 212 939 7004
Email: hgs+ecrit@cs.columbia.edu
URI: http://www.cs.columbia.edu

Appendix A. Configuring Emergency Dial Strings

Full Copyright Statement

Copyright (C) The section below is not directly related to the problem of
determining service location, but is an instance of the more generic
problem solved by this architecture, namely mapping location
information to URLs.

For the foreseeable future, some user devices and software will
emulate the user interface of a telephone, i.e., the only way to
enter call address information is via a 12-button keypad with digits
and the asterisk and hash symbol. Also, emergency numbers are likely
to be used until essentially all communication devices feature IP
connectivity and an alphanumeric keyboard. Unfortunately, more than
60 emergency numbers are in use throughout the world, with many of
those numbers serving non-emergency purposes elsewhere, e.g.,
identifying repair or directory services. Countries also
occasionally change their emergency numbers to conform to regional
agreements. An example is the introduction of 112 for countries in
Europe.

Thus, a system that allows devices to be used internationally to
place emergency calls needs to allow devices to discover emergency
numbers automatically. In the system proposed, these numbers are
strictly of local significance and are generally not visible in call
signaling messages.

For simplicity of presentation, this section assumes that emergency
numbers are valid throughout a country, rather than, say, be
restricted to a particular city. Internet Society (2006).

This appears likely to be true in
countries that have sufficiently advanced infrastructure to
contemplate deploying IP-based emergency calling solutions. In
addition, the solution proposed also works if certain countries do
not use a national emergency number. There document is no requirement that a
country uses a single emergency number for all emergency services,
such as fire, police, or rescue.

For the best user experience, systems should be able subject to discover two
sets of numbers, namely those used in the user's home country rights, licenses and restrictions
contained in
the country the user is currently visiting. The user is most likely
to remember the former, but a companion borrowing a device in an
emergency may only know the local emergency numbers.

Determining home and local emergency numbers is a configuration
problem, but unfortunately, existing configuration mechanisms are
ill-suited for this purpose. For example, a DHCP server might be
able to provide the local emergency number, but not the home numbers.
When virtual private networks (VPNs) are used, even DHCP may provide
numbers of uncertain origin, as a user may contact to the home
network or some local branch office of the corporate network.
Similarly, SIP configuration would be able to provide the numbers
valid at the location of the SIP service provider, but even a SIP
service provider with national footprint may serve customers that are
visiting any number of other countries.

Since dial strings are represented as URLs [5], the problem of
determining local BCP 78, and home emergency numbers is a problem of mapping
locations to a set of URLs, i.e., exactly the problem that the
mapping architecture is solving already.

The mapping operation is almost exactly the same except as for determining
the emergency service URL. The only difference is that if a seeker
knows the civic location at least to the country level, it will use a
query where the PIDF-LO only includes the country code. If it only
knows its geospatial location, it has to include that longitude and
latitude. The seeker uses the service identifiers "dialstring.sos",
"dialstring.sos.fire", etc. The resolver returns the appropriate set
of URLs and, if a geospatial location was used in the query, the
current region map for the country.

Within the mapping system, emergency calling regions are global
information, i.e., they are distributed using forth therein, the forest guide
replication mechanism described earlier. Thus, every forest guide
has access to authors
retain all region mappings. their rights.

This makes it possible that a
seeker can ask any resolver for this information, reducing the
privacy threat of revealing its location outside of an emergency
call. The privacy threat is further reduced by the long-lived nature
of the information, i.e., in almost all cases, the seeker will have
already cached document and the national boundary information or country
information on its first visit to the country.

Appendix B. Acknowledgments

Jong Yul Kim, Andrew Newton, Richard Stastny, Hannes Tschofenig contained herein are provided helpful comments. on an
"AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS
OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND THE INTERNET
ENGINEERING TASK FORCE DISCLAIM 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.

Intellectual Property Statement

The IETF takes no position regarding the validity or scope of any
Intellectual Property Rights or other rights that might be claimed to
pertain to the implementation or use of the technology described in
this document or the extent to which any license under such rights
might or might not be available; nor does it represent that it has
made any independent effort to identify any such rights. Information
on the procedures with respect to rights in RFC documents can be
found in BCP 78 and BCP 79.

Copies of IPR disclosures made to the IETF Secretariat and any
assurances of licenses to be made available, or the result of an
attempt made to obtain a general license or permission for the use of
such proprietary rights by implementers or users of this
specification can be obtained from the IETF on-line IPR repository at
http://www.ietf.org/ipr.

The IETF invites any interested party to bring to its attention any
copyrights, patents or patent applications, or other proprietary
rights that may cover technology that may be required to implement
this standard. Please address the information to the IETF at
ietf-ipr@ietf.org.

Disclaimer of Validity

This document and the information contained herein are provided on an
"AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS
OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND THE INTERNET
ENGINEERING TASK FORCE DISCLAIM 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.

Copyright (C) The Internet Society (2006). This document is subject
to the rights, licenses and restrictions contained in BCP 78, and
except as set forth therein, the authors retain all their rights.

Acknowledgment

Funding for the RFC Editor function is currently provided by the
Internet Society. IETF
Administrative Support Activity (IASA).