The CCNx protocol does not specify any assignment of meaning to names, only their structure. All assignment of meaning comes from application, institution, and/or global conventions. Application developers are free to design their own custom conventions. In order to have globally-routable names, institutional or global conventions must be reflected in prefix forwarding rules.

A number of basic issues requiring some convention will arise in many applications, and it will be helpful for everyone if common conventions are shared by many applications. In practice, such conventions should be implemented consistently in libraries.

This document specifies basic conventions implemented in the libraries of the CCNx Reference Implementation addressing the following issues:

  • Encoding of human-readable name components

  • Globally-unique name prefixes

  • Encoding of functional name components not intended to be directly human-readable

  • Naming of versions

  • Naming of segments (sequence numbering)

  • Naming file meta-data including headers

  • Functional/Command name components for constructing application-level protocols

Human-readable name components

Human readable clear-text components of names, analogous to components of file names, should use UTF-8 encoding and are always case-sensitive. In the case of encrypted names, the unencrypted values may be human-readable but the cipher-text names that appear in the messages will of course not be human-readable.

Globally-unique name prefixes

The first component of a globally-meaningful name should be a DNS name. Although the CCNx protocol does not rely on DNS, as a transition strategy it is helpful to take advantage of the existing DNS collection of globally-unique names. Furthermore, the DNS can be used to look up IP addresses for creating IP tunnels to provide CCNx protocol connectivity between sites.

It is not necessary to use globally-meaningful names exclusively. In some situations a local and context sensitive name prefix is appropriate. The specification of such contextual name prefixes is a matter of convention for particular applications. It is also unnecessary to use IP tunnels or other large-scale routing to communicate locally using globally-meaningful names.

Markers: Coding space for functional name components

In a variety of situations, name components need to represent commands or annotations that can appear at arbitrary levels in the name hierarchy, intermixed with the human readable components that are analogous to pieces of file names. In these cases, the position in the name is not able to distinguish the component’s role and so we rely on coding conventions in the component itself. Applications should follow these conventions so as to make the markers unambiguous. Wherever possible, however, the function or meaning of a name component should be determined by its context within the overall name, and human readable components should be used where they will not be mistaken for organizational name components.

The specification of UTF-8 prohibits certain octet values from occurring anywhere in a UTF-8 encoded string. These are the (hex) octet values C0, C1, and F5 to FF. In addition, the null octet 00 is used for string termination and so does not encode a usable character. Thus these make good markers to identify components of a CCNx name that play special functional roles such as for versioning and sequence numbering.

These conventions assign the prohibited octet values to specific functions. A functional name component is constructed using a marker octet from the UTF-8 prohibited set as the first octet, followed by additional octets according to the conventions for that marker.

Versions (Marker FD)

The first byte of a version component is 0xFD. Versions are named based (primarily) on time. The remaining bytes are a big-endian binary encoding of the time in CCNx Timestamp format.

The publisher generating the version stamp should try to avoid using a stamp earlier than (or the same as) any version of the file, to the extent that it knows about it. It should also avoid generating stamps that are unreasonably far in the future.

February 13, 2009 3:31:30 PM PST = 1234567890.000 seconds since January 1, 1970 UT
1234567890 = 0x499602d2
<Component ccnbencoding="hexBinary">FD 0 499602d2 000</Component>
<Component ccnbencoding="base64Binary">/QSZYC0gAA==</Component>

Versions, when present, commonly occupy the next-to-last component of the CCNx name, not counting the digest component. It is normal to have versions within versions, however, so version components MUST be interpreted contextually (see meta-data section).

The content within a version should include only segments and/or meta-data.

Segments (Markers 00, FB)

For consecutive numbering of segments, the first byte of the sequence component is 0x00. For non-consecutive numbering (e.g, using byte offsets) the value 0xFB should be used as the marker. In both cases, the remaining bytes of the segment name hold the sequence number in big-endian binary, using the minimum number of bytes. Thus sequence number 0 is encoded in just one byte, for example in consecutive numbering the name of segment 0 is %00 and the name of segment 1 is %00%01. The encoding of segments is typically limited to 9 bytes including the marker byte.

Note that this encoding is not quite dense - %00%00 is unused, as are other components that start with these two bytes.

Example: Sequence 257 is
<Component ccnbencoding="hexBinary">000101</Component>
<Component ccnbencoding="base64Binary">AAEB</Component>

Sequence numbers typically occupy the final component of the CCNx name (again, not counting the digest component).

Extensible Command Markers (Marker C1)

The marker 0xC1 prefixes namespaces for commands, queries, or other special identifiers, all of which will be referred to as commands in the following description.

Commands consist of the marker 0xC1 followed by ".", followed by a namespace, which is a UTF-8 string possibly containing additional periods, followed by an operation name which is a UTF-8 string interpreted by the namespace owner in whatever manner they choose. Since a namespace may have one or more interior periods so it may be constructed using a reversed DNS name, as with Java conventions for collision-free class naming.

Optionally the operation name may be followed by UTF-8 arguments delimited by tilde characters (~).

The UTF-8 arguments, if any, may be followed by a single, optional, binary argument. A general binary argument is prefixed by the byte 0x00. A binary argument that is ccnb-encoded data is prefixed by the byte 0xC1.

Note that early versions of the command marker syntax did not use a separator character between the operation and/or final UTF-8 argument and the last binary argument (previously treated as a data field). This lead to an ambiguous encoding parsable only by elements that knew the specific argument structure of a particular command. It is possible to encounter legacy data of this form stored in older repositories.

Example of General Command:

would be a command in the namespace org.ccnx, where the operation is frobnicate, which takes two arguments, in this case given as 1 and 37.

Namespaces containing only capital letters are reserved for CCNx standard application protocols. The protocols defined so far are listed below:

Repository Protocol (Namespace R)

The repository protocol is used for clients to control repositories which are stores of CCNx content, usually persistent stores.

Start Write Command

%C1.R.sw - request repository to retrieve and store content. The prefix of the name before this start write command component must be the prefix of the content to be written. Following this there should be a nonce component.

Checked Start Write Command

%C1.R.sw-c - similar to above, but the repository firsts checks for the presence of the stream and does not store it if it is already present. The repository responds in either case. Following the command component is a nonce component, a segment name component, and an explicit digest component, the last two identifying precisely the first (or only) segment of the stream to be stored.

Nonce (Namespace N)

The nonce namespace is used for marking random nonce values designed to make names unique.

Nonce Command

%C1.N%00<noncevalue> - where <noncevalue> is the random (binary) nonce value.

Early versions of this profile used an operation of 'n', and predated
the binary argument separator, resulting in nonce commands of

Name Enumeration (Namespace E)

The Name Enumeration Protocol is used to query for names of available content without retrieving the content itself.

Basic Enumeration Command - request first level names under the prefix of this command, much like a single-level directory listing command.

Marker (Namespace M)

The Marker namespace identifies name components meant to be parsed by programs, rather than people. They may contain additional components that allow them to be identified and parsed. They usually do not have command namespaces of their own because they are not operations, and do not have subcommands; but they are marked so that they can be easily filtered out from displays shown to the user and do not encounter accidental collisions with application names.

Marker namespaces in common use:

Guid name components:

These are used to specify fixed random binary name components, which are used repeatedly over time, rather than nonces.

%C1.M.G%00<guid value>

Key ID name components:

These refer to a particular public key by its digest (currently SHA256). Equivalent to PublisherPublicKeyDigest, and the PublisherID included in the SignedInfo of each ContentObject.

%C1.M.K%00<key id value>

Access control marker:

Content relating to access control for a given namespace node is stored below this marker component.


Scoped name:

This is used as the first component of a namespace that is restricted to the local machine.


Similarly, ‘%C1.M.S.neighorhood’ may be used as the first component of a namespace that is meant for our host and those directly connected. Note that, unlike ‘%C1.M.S.localhost’, this not rigorously enforced at present.


Service discovery:

This is used to denote a portion of the name tree used for service discovery.


For example /%C1.M.S.localhost/%C1.M.SRV/ccnd/KEY will get the local ccnd’s public key. Similarly for other services.


Meta-data content about files — headers, thumbnails, tags, and so on — is named in a subcollection of the file version named using the metadata namespace marker.


Metadata content is stored under this name component, named under human-readable names (as they are unlikely to collide with user-specified filenames).

A header is a CCN content stream that provides summary file-level information about the set of content to which it is attached by name. The name of a header is .header (again human-readable UTF-8). Note that a header is itself a file that has its own versions and segments within those.

Early versions of this profile used the name component meta for the metadata namespace, instead of %C1.META.