Handle System is the Corporation for National Research
Initiatives's proprietary registry assigning persistent identifiers,
or handles, to information resources, and for resolving "those handles
into the information necessary to locate, access, and otherwise make
use of the resources."
As with handles used elsewhere in computing,
Handle System handles are
opaque, and encode no information about the underlying resource, being
bound only to metadata regarding the resource. Consequently, the
handles are not rendered invalid by changes to the metadata.
The system was developed by
Bob Kahn at the Corporation for National
Research Initiatives (CNRI). The original work was funded by the
Defense Advanced Research Projects Agency
Defense Advanced Research Projects Agency (DARPA) between 1992 and
1996, as part of a wider framework for distributed digital object
services, and was thus contemporaneous with the early deployment of
the World Wide Web, with similar goals.
Handle System was first implemented in autumn 1994, and was
administered and operated by CNRI until December 2015, when a new
"multi-primary administrator" (MPA) mode of operation was introduced.
The DONA Foundation now administers the system's Global Handle
Registry and accredits MPAs, including CNRI and the International DOI
Foundation. The system currently provides the underlying
infrastructure for such handle-based systems as Digital Object
Identifiers and DSpace, which are mainly used to provide access to
scholarly, professional and government documents and other information
CNRI provides specifications and the source code for reference
implementations for the servers and protocols used in the system under
a royalty-free "Public License", similar to an open source license.
Thousands of handle services are currently running. Over 1000 of these
are at universities and libraries, but they are also in operation at
national laboratories, research groups, government agencies, and
commercial enterprises, receiving over 200 million resolution requests
3 Design principles
5 Licences and use policy
6 Related technologies
7 See also
9 External links
Handle System is defined in informational RFCs 3650, 3651
and 3652 of the
Internet Engineering Task Force
Internet Engineering Task Force (IETF); it includes
an open set of protocols, a namespace, and a reference implementation
of the protocols. Documentation, software, and related information is
provided by CNRI on a dedicated website
Handles consist of a prefix which identifies a "naming authority" and
a suffix which gives the "local name" of a resource. Similar to domain
names, prefixes are issued to naming authorities by one of the
"multi-primary administrators" of the system upon payment of a fee,
which must be renewed annually. A naming authority may create any
number of handles, with unique "local names", within their assigned
prefixes. An example of a handle is:
In the first example, which is the handle for the HANDLE.NET software
license, 20.1000 is the prefix assigned to the naming authority (in
this case, Handle.net itself) and 100 is the local name within that
namespace. The local name may consist of any characters from the
UCS-2 character set. The prefix also consists of any UCS-2
characters, other than "/". The prefixes consist of one or more naming
authority segments, separated by periods, representing a hierarchy of
naming authorities. Thus, in the example 20 is the naming authority
prefix for CNRI, while 1000 designates a subordinate naming authority
within the 20 prefix. Other examples of top-level prefixes for the
federated naming authorities of the DONA Foundation are 10 for DOI
handles; 11 for handles assigned by the ITU; 21 for handles issued by
the German Gesellschaft für wissenschaftliche Datenverarbeitung mbH
Göttingen (GWDG), the scientific computing center of the University
of Göttingen; and 86 for the Coalition of Handle Services — China.
Older "legacy" prefixes issued by CNRI before the "multi-primary
administrator" (MPA) structure was instituted are typically four of
five digits, as in the second example above, a handle administered by
the University of Leicester. All prefixes must be registered in the
Global Handle Registry through an DONA Foundation approved registrar,
normally for a fee.
As with other uses of handles in computing, the handle is opaque; that
is, it encodes no information about the underlying resource and
provides only the means to retrieve metadata about the resource.
This may be contrasted with a
Uniform Resource Locator (URL), which
may encode within the identifier such attributes of the resource as
the protocol to be used to access the server holding the resource, the
server host name and port number, and perhaps even location specifics
such as the name of a file in the server file system containing the
resource. In the Handle System, these specifics are not encoded in the
handle, but are found in the metadata to which the handle is bound.
The metadata may include many attributes of the information resource,
such as its locations, the forms in which it is available, the types
of access (e.g. "free" versus "paid") offered, and to whom. The
processing of the metadata to determine how and where the resource
should be accessed, and the provision of the resource to the user, are
performed in a separate step, called "resolution", using a Resolver, a
server which may be different than the ones involved in exchanging the
handle for the metadata. Unlike URLs, which may become invalid if the
metadata embedded within them becomes invalid, handles do not become
invalid and do not need to change when locations or other metadata
attributes change. This helps to prevent link rot, as changes in the
information resource (such as location) need only be reflected in
changes to the metadata, rather than in changes in every reference to
Each handle may have its own administrator(s) and administration of
the handles can be done in a distributed environment, similar to DNS
domain names. The name-to-value bindings may also be secured, both via
signatures to verify the data and via challenge response to verify the
transmission of the data, allowing handles to be used in trust
It is possible for the same underlying information resource to be
associated with multiple handles, as when two university libraries
generate handles (and therefore possibly different sets of metadata)
for the same book.
Handle System is compatible with the
Domain Name System
Domain Name System (DNS), but
does not require it, unlike persistent identifiers such as PURLs or
ARKs, which are similar to handles, but which utilise domain names.
However, unlike these domain-name based approaches, handles do require
a separate prefix registration process and handle servers separate
from the domain name servers.
Handles can be used natively. or expressed as Uniform Resource
Identifiers (URIs) through a namespace within the info URI scheme;
 for example, 20.1000/100 may be written as the URI,
Handle System namespaces, such as Digital
Object Identifiers, are "info:" URI namespaces in their own right; for
example, info:doi/10.1000/182 is another way of writing the handle for
the current revision of the DOI Handbook as a URI.
Handle System namespaces define special presentation rules. For
example, Digital Object Identifiers, which represent a high percentage
of the extant handles, are usually presented with a "doi:" prefix:
Handles may also be expressed as Uniform Resource Locators (URLs)
through the use of a
HTTP proxy server, such as:
Implementation of the
Handle System consists of Local Handle Services,
each of which is made up of one or more sites that provide the servers
that store specific handles. The Global Handle Registry is a unique
Local Handle Service which stores information on the prefixes (also
known as naming authorities) within the
Handle System and can be
queried to find out where specific handles are stored on other Local
Handle Services within this distributed system.
Handle System website provides a series of implementation tools,
notably the HANDLE.NET Software and HANDLE.NET Client
Libraries. Handle clients can be embedded in end user software
(e.g., a web browser) or in server software (e.g., a web server) and
extensions are already available for Adobe Acrobat and
Handle client software libraries are available in both C and Java.
Some applications have developed specific add-on tools, e.g., for the
The interoperable network of distributed handle resolver servers (also
known as the Proxy Server System) are linked through a Global Resolver
(which is one logical entity though physically decentralised and
mirrored). Users of
Handle System technology obtain a handle prefix
created in the Global Handle Registry. The Global Handle Registry
maintains and resolves the prefixes of locally maintained handle
services. Any local handle service can, therefore, resolve any handle
through the Global Resolver.
Handles (identifiers) are passed by a client, as a query of the naming
authority/prefix, to the Handle System's Global Handle Registry (GHR).
The GHR responds by sending the client the location information for
the relevant Local Handle Service (which may consist of multiple
servers in multiple sites); a query is then sent to the relevant
server within the Local Handle Service. The Local Handle Service
returns the information needed to acquire the resource, e.g., a URL
which can then be turned into an HTTP re-direct. (Note: if the client
already has information on the appropriate LHS to query, the initial
query to GHR is omitted)
Though the original model from which the
Handle System derives dealt
with management of digital objects, the
Handle System does not mandate
any particular model of relationships between the identified entities,
nor is it limited to identifying only digital objects: non-digital
entities may be represented as a corresponding digital object for the
purposes of digital object management. Some care is needed in the
definition of such objects and how they relate to non-digital
entities; there are established models that can aid in such
definitions (e.g., Functional Requirements for Bibliographic Records
(FRBR), CIDOC CRM, and indecs content model. Some applications have
found it helpful to marry such a framework to the handle application:
for example, the Advanced Distributed Learning (ADL) Initiative
Handle System application with existing standards for
distributed learning content, using a Shareable Content Object
Reference Model (SCORM), and the
Digital Object Identifier
Digital Object Identifier (DOI)
system implementation of the
Handle System has adopted it together
with the indecs framework to deal with semantic interoperability.
Handle System also makes explicit the importance of organizational
commitment to a persistent identifier scheme, but does not mandate one
model for ensuring such commitment. Individual applications may choose
to establish their own sets of rules and social infrastructure to
ensure persistence (e.g., when used in the
DSpace application, and the
The Handle system is designed to meet the following requirements to
contribute to persistence
The identifier string:
is not based on any changeable attributes of the entity (location,
ownership, or any other attribute that may change without changing the
is opaque (preferably a ‘dumb number’: a well known pattern
invites assumptions that may be misleading, and meaningful semantics
may not translate across languages and may cause trademark conflicts);
is unique within the system (to avoid collisions and referential
has optional, but nice to have, features that should be supported
(human-readable,cut-and-paste-able, embeddable; fits common systems,
e.g., URI specification).
The identifier resolution mechanism:
is reliable (using redundancy, no single points of failure, and fast
enough to not appear broken);
is scalable (higher loads simply managed with more computers);
is flexible (can adapt to changing computing environments; useful to
is trusted (both resolution and administration have technical trust
methods; an operating organization is committed to the long term);
builds on open architecture (encouraging the leverage efforts of a
community in building applications on the infrastructure);
is transparent (users need not know the infrastructure details).
Among the objects that are currently identified by handles are journal
articles, technical reports, books, theses and dissertations,
government documents, metadata, distributed learning content, and data
sets. Handles are being used in digital watermarking applications,
GRID applications, repositories, and more. Although individual users
may download and use the HANDLE.NET software independently, many users
have found it beneficial to collaborate in developing applications in
a federation, using common policy or additional technology to provide
shared services. As one of the first persistent identifier schemes,
Handle System has been widely adopted by public and private
institutions and proven over several years. (See Paradigm, Persistent
Handle System applications may use handles as simple persistent
identifiers (as most commonly used, to resolve to the current URL of
an object), or may choose to take advantage of other features. Its
support for the simultaneous return as output of multiple pieces of
current information related to the object, in defined data structures,
enables priorities to be established for the order in which the
multiple resolutions will be used. Handles can, therefore, resolve to
different digital versions of the same content, to mirror sites, or to
different business models (pay vs. free, secure vs. open, public vs.
private). They can also resolve to different digital versions of
differing content, such as a mix of objects required for a
There are thousands of handle services running today, located in 71
countries, on 6 continents; over 1000 of them run at universities and
libraries. Handle services are being run by user federations, national
laboratories, universities, computing centers, libraries (national and
local), government agencies, contractors, corporations, and research
groups. Major publishers use the
Handle System for persistent
identification of commercially traded and Open Access content through
its implementation with the
Digital Object Identifier
Digital Object Identifier (DOI) system.
The number of prefixes, which allow users to assign handles, is
growing and stands at over 12,000 as of early 2014. There are six
top-level Global Handle Registry servers that receive (on average) 68
million resolution requests per month. Proxy servers known to CNRI,
passing requests to the system on the Web, receive (on average) 200
million resolution requests per month. (Statistics from Handle Quick
In 2010, CNRI and
ITU (International Telecommunication Union) entered
into an agreement to collaborate on use of the
Handle System (and the
Digital Object Architecture more generally) and are working on the
specific details of that collaboration; in April 2009
ITU listed the
Handle System as an "emerging trend".
Licences and use policy
Handle System, HANDLE.NET and Global Handle Registry are trademarks of
Corporation for National Research Initiatives (CNRI), a non-profit
research and development corporation in the USA. The
Handle System is
the subject of patents by CNRI, which licenses its Handle System
technology through a public license, similar to an open source
license, in order to enable broader use of the technology. Handle
System infrastructure is supported by prefix registration and service
fees, with the majority coming from single prefix holders. The largest
current single contributor is the International DOI Foundation. The
Public License allows commercial and non-commercial use at low cost of
both its patented technology and the reference implementation of the
software, and allows the software to be freely embedded in other
systems and products. A Service Agreement is also available for
users who intend to provide identifier and/or resolution services
Handle System technology under the
Handle System public
Handle System represents several components of a long-term digital
object architecture. In January 2010 CNRI released its general-purpose
Digital Object Repository software, another major component of
this architecture. More information about the release, including
protocol specification, source code and ready-to-use system, clients
and utilities, is available.
Resource Description Framework
Uniform Resource Name
Uniform Resource Identifier
Archival Resource Key
^ a b "RFC 3650:
Handle System Overview".
^ "Kahn/Wilensky Architecture". CNRI. 1995-05-13. Retrieved
^ "Welcome to The DONA Foundation".
^ "The DOI System".
^ a b "Redirect to Current Handle.Net web site content".
www.handle.net. Retrieved 15 March 2018.
^ "RFC 3651:
Handle System Namespace and Service Definition".
^ "RFC 3652:
Handle System Protocol (ver 2.1) Specification".
^ "handle.net". handle.net. Retrieved 2013-03-13.
^ "About "info" URIs - Frequently Asked Questions". Info-uri.info.
^ "RFC 4452: The "info" URI Scheme for Information Assats with
Identifiers in Public Namespaces".
^ "DOI Handbook".
^ "HDL.NET Services: Proxy Server System". Handle.net. Retrieved
^ "HS Software Download". Handle.net. Retrieved 2013-03-13.
^ "Software Client Libraries". Handle.net. Retrieved 2013-03-13.
^ "HDL Plug-in for
Adobe Acrobat and Acrobat Reader". Handle.net.
^  Archived September 5, 2015, at the Wayback Machine.
^ "DOI System Tools". Doi.org. 2012-07-12. Retrieved 2013-03-13.
^ "adlnet.gov". adlnet.gov. Retrieved 2013-03-13.
^ "SCORM". adlnet.gov. Archived from the original on 2008-06-06.
^ "doi.org". doi.org. 2013-01-08. Retrieved 2013-03-13.
^ "Identifier Systems in Network Architecture, Laurence Lannom, CNRI.
Video of presentation (or presentation PDF only) from the Digital
Metadata Symposium, Science & Technology Council,
Academy of Motion Picture Arts & Sciences, 11 June 2009".
Oscars.org. 2012-08-24. Archived from the original on 2013-03-30.
^ "workbook on digital private papers administrative and
preservation metadata persistent identifiers". paradigm. 2008-01-02.
^ "Handle System". Itu.int. 2010-04-16. Retrieved 2013-03-13.
^ "dorepository.org". dorepository.org. 2013-01-08. Retrieved
^ "Digital Object Repository Server: A Component of the Digital Object
Architecture". Dlib.org. 2010-02-04. Retrieved 2013-03-13.
^ "DO Repository". DO Repository. doi:10.1045/january2010-reilly.
Persistent identifiers p