Network News Transport Protocol

This document specifies the Network News Transport Protocol (NNTP), which is used for the distribution, inquiry, retrieval, and posting of Netnews articles using a reliable stream-based mechanism. For news reading clients, NNTP enables retrieval of news articles that are stored in a central database, giving subscribers the ability to select only those articles they wish to read. The Netnews model provides for indexing, cross-referencing, and expiration of aged messages. For server-to-server interaction, NNTP is designed for efficient transmission of Netnews articles over a reliable full duplex communication channel. Every attempt is made to ensure that the protocol specification in this document is compatible with the version specified in RFC 977. However, this version does not support the ill-defined SLAVE command and permits four digit years to be specified in the NEWNEWS and NEWGROUPS commands. It changes the default character set to UTF-8 instead of US-ASCII. It now requires all articles to have a message-id, eliminating the "<0>" placeholder used in RFC 977. It also extends the newsgroup name matching capabilities already documented in RFC 977. Generally, new functionality is made available using new commands. Part of that new functionality involves a mechanism to discover what new functionality is available to clients from a server. This mechanism can also be used to add more functionality as needs merit such additions. The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119. An implementation is not compliant if it fails to satisfy one or more of the MUST requirements for this protocol. An implementation that satisfies all the MUST and all the SHOULD requirements for its protocols is said to be "unconditionally compliant"; one that satisfies all the MUST requirements but not all the SHOULD requirements for NNTP is said to be "conditionally compliant". For the remainder of this document, the term "client" or "client host" refers to a host making use of the NNTP service, while the term "server" or "server host" refers to a host that offers the NNTP service.

The following notational conventions are used in this document.

UPPERCASE indicates literal text to be included in the command; lowercase indicates a token described elsewhere; [brackets] indicate that the parameter is optional; ellipsis... indicates that the parameter may be repeated any number of times (it must occur at least once); vertical|bar indicates a choice of two mutually exclusive parameters (exactly one must be provided). The name "message-id" for a command or response parameter indicates that it is the message-id of an article as described in , including the angle brackets. The name "wildmat" for a parameter indicates that it is a wildmat as defined in . If the parameter does not meet the requirements of that section (for example, if it does not fit the grammar of ) the NNTP server MAY place some interpretation on it (not specified by this document) or otherwise MUST treat it as a syntax error. Responses for each command will be described in tables listing the required format of a response followed by the meaning that should be ascribed to that response. The terms "NUL", "TAB", "LF", "CR, and "space" refer to the octets with those codes in US-ASCII (that is, %x00, %x09, %x0A, %x0D, and %x20 respectively), as do quoted characters (so "." and "<" refer to %x2E and %x3C). The term "CRLF" or "CRLF pair" means the sequence CR immediately followed by LF (that is, %x0D.0A). A "printable US-ASCII character" is an octet in the range %x21-7E. Examples in this document are not normative but serve to illustrate usages, arguments, and responses. In the examples, a "[C]" will be used to represent the client host and a "[S]" will be used to represent the server host. Most of the examples do not rely on a particular server state. In some cases, however, they do assume that the current selected newsgroup (see the GROUP command) is invalid; when so, this is indicated at the start of the example.

NNTP operates over any reliable data stream 8-bit-wide channel. Initially, the server host starts the NNTP service by listening on a TCP port; when running over TCP/IP, the official port for the NNTP service is 119. When a client host wishes to make use of the service, it MUST establish a TCP connection with the server host by connecting to that host on the same port on which the server is listening. When the connection is established, the NNTP server host MUST send a greeting. The client host and server host then exchange commands and responses (respectively) until the connection is closed or aborted. The character set for all NNTP commands is UTF-8. Commands in NNTP MUST consist of a keyword, which MAY be followed by one or more arguments. A CRLF pair MUST terminate all commands. Multiple commands MUST NOT be on the same line. Keywords MUST consist of printable US-ASCII characters. Unless otherwise noted elsewhere in this document, arguments SHOULD consist of printable US-ASCII characters. Keywords and arguments MUST be each separated by one or more space or TAB characters. Keywords MUST be at least three characters and MUST NOT exceed 12 characters. Command lines MUST NOT exceed 512 octets, which includes the terminating CRLF pair. The arguments MUST NOT exceed 497 octets. Where this specification permits UTF-8 characters outside the range U+0000 to U+007F, implementations MUST NOT use the Byte Order Mark (U+FEFF, encoding %xEF.BB.BF), and MUST use the Word Joiner (U+2060, encoding %xE2.91.A0) for the meaning Zero Width No-Break Space, in command lines and the initial lines of responses, and SHOULD apply these same principles throughout. Commands may have variants, using a second keyword immediately after the first to indicate which variant is required. The only such commands in this specification are LIST and MODE. Keywords are case-insensitive; the case of keywords for commands MUST be ignored by the server. Command and response parameters are case or language specific only when stated, either in this document or in other relevant specifications. An NNTP server MUST implement all the commands in this specification except for those marked as optional and those in extensions. Each response MUST start with a three-digit response code that is sufficient to distinguish all responses. Certain valid responses are defined to be multi-line; for all others, the response is contained in a single line. Should the initial response line be limited to 512 octets as well? Possible text: The first or only line of the response MUST NOT exceed 512 octets, which includes the response code and the terminating CRLF pair. The text further down about "does not place any limit on the length" would need equivalent edits. All multi-line responses MUST adhere to the following format: The response consists of a sequence of one or more "lines", each being a stream of octets ending with a CRLF pair. Apart from those line endings, the stream MUST NOT include the octets NUL, LF, or CR. The first such line contains the response code as with a single line response. If any subsequent line begins with the "termination octet" ("." or %x2E), that line MUST be "byte-stuffed" by pre-pending an additional termination octet to that line of the response. The lines of the response MUST be followed by a terminating line consisting of a single termination octet followed by a CRLF pair in the normal way. Thus a multi-line response is always terminated with the five octets CRLF "." CRLF (%x0D.0A.2E.0D.0A). When interpreting a multi-line response, the "byte stuffing" MUST be undone; i.e. the client MUST ensure that, in any line beginning with the termination octet followed by octets other than a CRLF pair, that initial termination octet is disregarded. Likewise, the terminating line ("." CRLF or %x2E.0D.0A) MUST NOT be considered part of the multi-line response; i.e. the client MUST ensure that any line beginning with the termination octet followed immediately by a CRLF pair is disregarded; (the first CRLF pair of the terminating CRLF "." CRLF is, of course, part of the last line of the response). Note that texts using an encoding (such as UTF-16 or UTF-32) that may contain the octets NUL, LF, or CR other than a CRLF pair cannot be reliably conveyed in the above format. However, except when stated otherwise, this specification does not require the content to be UTF-8 and it is possible for octets above and below 128 to be mixed arbitrarily. This document does not place any limit on the length of a line. However, the standards that define the format of articles may do so. An NNTP server MAY have an inactivity autologout timer. Such a timer SHOULD be of at least three minutes duration, with the exception that there MAY be a shorter limit on how long the server is willing to wait for the first command from the client. The receipt of any command from the client during the timer interval SHOULD suffice to reset the autologout timer. Similarly, the receipt of any significant amount of data from the client while in the midst of sending a multi-line message to the server (such as during a POST or IHAVE command) SHOULD suffice to reset the autologout timer. When the timer expires, the server SHOULD close the TCP connection without sending any response to the client.

Each response MUST begin with a three-digit status indicator. These are status reports from the server and indicate the response to the last command received from the client. The first digit of the response broadly indicates the success, failure, or progress of the previous command. 1xx - Informative message. 2xx - Command completed OK. 3xx - Command OK so far; send the rest of it. 4xx - Command was correct, but couldn't be performed for some reason. 5xx - Command unimplemented, or incorrect, or a serious program error occurred. I proposed that we assign 6xx for extensions and future commands to use for multiline responses, thus at least limiting (if not eliminating) the problem clients have of working out whether one is coming up. Nobody was violently against the idea, but nobody was particularly in favour either. The next digit in the code indicates the function response category. x0x - Connection, setup, and miscellaneous messages x1x - Newsgroup selection x2x - Article selection x3x - Distribution functions x4x - Posting x8x - Reserved for authentication and authorization extensions x9x - Reserved for private use (non-standard extensions) Certain responses contain parameters such as numbers and names in addition to the status indicator. In those cases, to simplify interpretation by the client the number and type of such parameters is fixed for each response code, as is whether or not the code introduces a multi-line response. Any extension MUST follow this principle as well, but note that, for historical reasons, the 211 response code is an exception to this. In all other cases, the client MUST only use the status indicator itself to determine the nature of the response. The exact response codes that can be returned by any given command are detailed in the description of that command. Parameters MUST be separated from the numeric status indicator and from each other by a single space. All numeric parameters MUST be in base 10 (decimal) format, and MAY have leading zeros. String parameters MUST contain at least one character and MUST NOT contain TAB, LF, CR, or space. The server MAY add any text after the response code or last parameter as appropriate, and the client MUST NOT make decisions based on this text. Such text MUST be separated from the numeric status indicator or the last parameter by at least one space. The server MUST respond to any command with the appropriate generic response (given in ) if it represents the situation. Otherwise, each recognized command MUST return one of the response codes specifically listed in its description or in an extension. A server MAY provide extensions to this specification, including new commands, new variants or features of existing commands, and other ways of changing the internal state of the server. However, the server MUST NOT produce any other responses to a client that does not invoke any of the additional features. (Therefore a client that restricts itself to this specification will only receive the responses that are listed.) If a client receives an unexpected response, it SHOULD use the first digit of the response to determine the result. For example, an unexpected 2xx should be taken as success and an unexpected 4xx or 5xx as failure. Response codes not specified in this document MAY be used for any installation-specific additional commands also not specified. These SHOULD be chosen to fit the pattern of x9x specified above. Neither this document nor any extension registered with IANA (see ) will specify any response codes of the x9x pattern. (Implementers of extensions are accordingly cautioned not to use such responses for extensions that may subsequently be submitted for registration.)

The server MUST respond to any command with the appropriate one of the following generic responses if it represents the situation. If the command is not recognized, or it is an optional command or extension that is not implemented by the server, the response code 500 MUST be returned. If there is a syntax error in the arguments of a recognized command, including the case where more arguments are provided than the command specifies, the response code 501 MUST be returned. Note that where a command has variants depending on a second keyword (e.g. LIST ACTIVE and LIST NEWSGROUPS), then 501 MUST be used when the requested variant is not implemented but the base command is. If the server experiences an internal fault or problem that means it is unable to carry out the command (for example, a necessary file is missing or a necessary service could not be contacted), the response code 403 MUST be returned. If the server recognises the command but does not provide an optional feature (for example because it does not store the required information), or only handles a subset of legitimate cases (see the HDR command for an example), the response code 503 MUST be returned. Note that where a command is optional (e.g. LIST ACTIVE.TIMES) and is not provided by a server, this MAY be treated as an unimplemented command (response code 500 or 501) or as a working command where the information is not available (response code 503). Do we need to add text like: For backwards compatibility a server MAY return the response code 503 where this specification requires the response code 403, and a client SHOULD be prepared for this. This waiver may be removed in a future revision of this specification. If the client is not authorized to use the specified facility when the server is in its current state, then either the response code 480 or the response code 502 MUST be returned. The response code 480 SHOULD be used if a different command (for example, an extension used to present credentials) might change the server state so that the command is permitted. The response code 502 SHOULD be used if the server wishes to indicate that it is necessary to terminate the connection and start a new one with the appropriate authority before the command can be used. Since it is not always possible to clearly distinguish these two cases, a server MAY issue either of these response codes for either case. (Note that the server MUST NOT close the TCP connection immediately after a 502 response except at the initial connection and with the MODE READER command.) This isn't a complete solution to the 480 issue; what about the TLS extension, which uses 483 to mean "you need encryption". Should 480 be used for other than "you need authentication"? What code should be used to mean "can't do AUTH until after MODE READER"? Do we need a more generic mechanism for "you must invoke extension X to do Y"? The best proposal made so far is that all 48x codes, if returned from an existing command, mean "unavailable unless some authentication or privacy extension is invoked". Does this tie in with the issue of permitting existing commands not listed in an extension? If the server has to terminate the connection for some reason, it MUST give a 400 response code to the next command and then immediately close the TCP connection. It MAY give a 401 response code to any command to indicate that termination is imminent (following a 401 response, it MUST NOT close the TCP connection immediately). It's not clear that we need 401; it appears to have been an invention. If we do keep it, then text is needed to indicate what happens with commands that change the status (for example, if GROUP returns 401 what happens to the current selected newsgroup), and how to make those commands work. With the exception of mandatory commands and the 500 response, the client MUST be prepared to receive any of these responses for any command. Example of an unknown command: Example of an unsupported extension: Example of an unsupported variant: Example of a syntax error: Example of an overlong command line: Example of a bad wildmat: Example of an attempt to access a restricted facility: followed by a successful attempt following authentication: Example of an attempt to access a facility not available to this connection: Example of a temporary failure: Example of the server needing to close down immediately: Example of imminent termination of the server:

NNTP is designed to operate over a reliable bi-directional connection such as TCP. Therefore, if a command does not depend on the response to the previous one, it should not matter if it is sent before that response is received. Doing this is called "pipelining". However, certain server implementations throw away all text received from the client following certain commands before sending their response. If this happens, pipelining will be affected because one or more commands will have been ignored or misinterpreted, and the client will be matching the wrong responses to each command. Since there are significant benefits to pipelining, but also circumstances where it is reasonable or common for servers to behave in the above manner, this document puts certain requirements on both clients and servers. Except where stated otherwise, a client MAY use pipelining. That is, it may send a command before receiving the response for the previous command. The server MUST allow pipelining and MUST NOT throw away any text received after a command. Irrespective of whether or not pipelining is used, the server MUST process commands in the order they are sent. If the specific description of a command says it "MUST NOT be pipelined", that command MUST end any pipeline of commands. That is, the client MUST NOT send any following command until receiving the CRLF at the end of the response from the command. The server MAY ignore any data received after the command and before the CRLF at the end of the response is sent to the client. The initial connection must not be part of a pipeline; that is, the client MUST NOT send any command until receiving the CRLF at the end of the greeting. If the client uses blocking system calls to send commands, it MUST ensure that the amount of text sent in pipelining does not cause a deadlock between transmission and reception. The amount of text involved will depend on window sizes in the transmission layer, and is typically 4k octets for TCP. Example of correct use of pipelining: Example of incorrect use of pipelining (the MODE READER command may not be pipelined): The DATE command has been thrown away by the server and so there is no 111 response to match it.

This section is new. If anyone has better wording, I won't complain. NNTP is intended to transfer articles between clients and servers. For the purposes of this specification, articles are required to conform to the rules in this section and clients and servers MUST correctly process any article received from the other that does so. Note that this requirement applies only to the contents of communications over NNTP; it does not prevent the client or server from subsequently rejecting an article for reasons of local policy. In particular, where NNTP is used to transport articles that conform to other specifications such as RFC 1036 or RFC 2822, articles must meet both this specification and that other. Need to add an appendix that spells out how this document interacts with RFC 1036. That would allow us to remove some of the convoluted wording about "other specifications". An article consists of two parts: the headers and the body. They are separated by a single empty line, or in other words by two consecutive CRLF pairs (if there is more than one empty line, the second and subsequent ones are part of the body). In order to meet the general requirements of NNTP, an article MUST NOT include the octet NUL, MUST NOT contain the octets LF and CR other than as part of a CRLF pair, and MUST end with a CRLF pair. This specification puts no further restrictions on the body; in particular, it MAY be empty. The headers of an article consist of one or more header lines. Each header line consists of a header name, a colon, a space, the header content, and a CRLF in that order. The name consists of one or more printable US-ASCII characters other than colon and, for the purposes of this specification, is not case sensitive. There MAY be more than one header line with the same name. The content MUST NOT contain CRLF but is otherwise unrestricted; in particular, it MAY be empty. A header may be "folded"; that is, a CRLF pair may be placed before any TAB or space in the line (including the space after the colon after the header name), except that there MUST be at least one octet other than %x09 or %x20 between any two CRLF pairs in a header line. (Note that folding means that the header line occupies more than one line when displayed or transmitted; nevertheless it is still referred to as "a" header line.) The presence or absence of folding does not affect the meaning of the header line; that is, the CRLF pairs introduced by folding are not considered part of the header value. Each article MUST have a unique message-id; two articles offered by an NNTP server MUST NOT have the same message-id. Note that RFC 1036 further requires that message-ids are globally unique for all time. For the purposes of this specification, message-ids are opaque strings that MUST meet the following requirements: A message-id MUST begin with "<" and end with ">", and MUST NOT contain the latter except at the end. A message-id MUST be between 3 and 250 octets in length. A message-id MUST NOT contain octets other than printable US-ASCII characters. Two message-ids are the same if and only if they consist of the same sequence of octets. Other specifications may define two different sequences as being equal; an NNTP server that also conforms to such a specification must consistently use only one or the other. As an example, the message-ids: <abcd@example.com> <"abcd"@example.com> <"ab\cd"@example.com> are considered distinct by this specification even though they would be considered semantically identical according to the specification in RFC 2822. This specification does not describe how the message-id of an article is determined (if the server is also conforming to another specification that contains a definition of message-id compatible with this one, the server SHOULD use those message-ids). Many servers will extract the message-id from the contents of a header with name "Message-ID", but this is not required by this document. If the server does not have any way to determine a message-id from the article itself, it MUST synthesise one (it need not modify the article to add such a header unless required to by another specification).

The WILDMAT format described here is based on the version first developed by Rich Salz , which in turn was derived from the format used in the UNIX "find" command to articulate file names. It was developed to provide a uniform mechanism for matching patterns in the same manner that the UNIX shell matches filenames.

A wildmat is described by the following ABNF syntax (note that this syntax contains ambiguities and special cases described at the end): wildmat-pattern *("," ["!"] wildmat-pattern) 1*wildmat-item wildmat-exact / wildmat-wild %x21-29 / %x2B / %x2D-3E / %x40-5A / %x5E-7E / UTF8-non-ascii ; exclude * , ? [ \ ] "*" / "?" UTF8-non-ascii is defined in . This syntax must be interpreted subject to the following rule: Where a wildmat-pattern is not immediately preceded by "!", it shall not begin with a "!". Note: the characters \ , [ and ] are not allowed in wildmats, while * and ? are always wildcards. This should not be a problem since these characters cannot occur in newsgroup names, which is the only current use of wildmats. Backslash is commonly used to suppress the special meaning of characters while brackets are used to introduce sets. However, these usages are not universal and interpretation of these characters in the context of UTF-8 strings is both potentially complex and differs from existing practice, so they were omitted from this specification. A future extension to this specification may provide semantics for these characters.

A wildmat is tested against a string, and either matches or does not match. To do this, each constituent wildmat-pattern is matched against the string and the rightmost pattern that matches is identified. If that wildmat-pattern is not preceded with "!", the whole wildmat matches. If it is preceded by "!", or if no wildmat-pattern matches, the whole wildmat does not match. For example, consider the wildmat "a*,!*b,*c*": the string "aaa" matches because the rightmost match is with "a*" the string "abb" does not match because the rightmost match is with "*b" the string "ccb" matches because the rightmost match is with "*c*" the string "xxx" does not match because no wildmat-pattern matches A wildmat-pattern matches a string if the string can be broken into components, each of which matches the corresponding wildmat-item in the pattern; the matches must be in the same order, and the whole string must be used in the match. The pattern is "anchored"; that is, the first and last characters in the string must match the first and last item respectively (unless that item is an asterisk matching zero characters). A wildmat-exact matches the same character (which may be more than one octet in UTF-8). "?" matches exactly one character (which may be more than one octet). "*" matches zero or more characters. It can match an empty string, but it cannot match a subsequence of a UTF-8 sequence that is not aligned to the character boundaries.

An NNTP server or extension MAY extend the syntax or semantics of wildmats provided that all wildmats that meet the requirements of have the meaning ascribed to them by . Future editions of this document may also extend wildmats.

In these examples, $ and @ are used to represent the two octets %xC2 and %xA3 respectively; $@ is thus the UTF-8 encoding for the pound sterling symbol, shown as # in the descriptions. Wildmat Description of strings that match abc the one string "abc" abc,def the two strings "abc" and "def" $@ the one character string "#" a* any string that begins with "a" a*b any string that begins with "a" and ends with "b" a*,*b any string that begins with "a" or ends with "b" a*,!*b any string that begins with "a" and does not end with "b" a*,!*b,c* any string that begins with "a" and does not end with "b", and any string that begins with "c" no matter what it ends with a*,c*,!*b any string that begins with "a" or "c" and does not end with "b" ?a* any string with "a" as its second character ??a* any string with "a" as its third character *a? any string with "a" as its penultimate character *a?? any string with "a" as its antepenultimate character

These are the only valid response codes for the initial greeting; the server MUST not return any other generic response code. Following a 400 or 502 response the server MUST immediately close the connection. There is no command presented by the client upon initial connection to the server. The server MUST present an appropriate response code as a greeting to the client. This response informs the client whether service is available and whether the client is permitted to post. If the server will accept further commands from the client including POST, the server MUST present a 200 greeting code. If the server will accept further commands from the client, but it is not authorized to post articles using the POST command, the server MUST present a 201 greeting code. Otherwise the server MUST present a 400 or 502 greeting code and then immediately close the connection. 502 MUST be used if the client is not permitted under any circumstances to interact with the server and 400 otherwise. Example of a normal connection from an authorized client which then terminates the session (see ): Example of a normal connection from an authorized client that is not permitted to post; it also immediately terminates the session: Example of a normal connection from an unauthorized client: Example of a connection from a client where the server is unable to provide service: Following a 400 or 502 response the server MUST immediately close the connection. MODE READER SHOULD be sent by any client that intends to use any command other than IHAVE, HEAD, STAT, LIST ACTIVE, LIST EXTENSIONS, or a command advertised by the server as available via LIST EXTENSIONS. Servers MAY require that this command be issued before any commands other than the above are sent and MAY reject such commands until after a MODE READER command has been sent. Where an extension is only available after a MODE READER command, or where the effects of the extension will change, the LIST EXTENSIONS command MUST produce different results that indicate the change. The server MUST return a response using the same codes as the initial greeting (as described in ) to indicate its ability to provide reading service to the client. Note that the response need not be the same as the one presented during the initial greeting. Once MODE READER is sent, IHAVE (and any extensions intended for peer-to-peer article transfer) MAY no longer be permitted, even if it were permitted before the MODE READER command. The results of LIST EXTENSIONS MAY be different following a MODE READER command than prior to the issuing of that command. Servers are encouraged to not require this command even though clients SHOULD send it when appropriate. It is present to support some news architectures that switch between modes based on whether a given connection is a peer-to-peer connection with another server or a news reading client. Example of use of the MODE READER command by an authorized client which then terminates the session (see ): Example of use of the MODE READER command by an authorized client that is not permitted to post; it also immediately terminates the session: Example of use of MODE READER by a client not authorized to receive service from the server as a news reader: Example of a connection from any client where the server is temporarily unable to provide news reader service: The LIST EXTENSIONS command allows a client to determine which extensions are supported by the server at any given time. See for further discussion of extensions. This command MUST be implemented by any server that implements any extensions defined in this document or any other extension in the IANA registry, and is optional otherwise. This command MAY be issued at anytime during a session. It is not required that the client issues this command before attempting to make use of any extension. The response generated by this command MAY change during a session because of other state information (which in turn may be changed by the effects of other commands). An NNTP client MUST NOT cache (for use in another session) any information returned if the LIST EXTENSIONS command succeeds. That is, an NNTP client is only able to get the current and correct information concerning available extensions at any point during a session by issuing a LIST EXTENSIONS command at that point of that session and processing the response. The list of extensions is returned as a multi-line response following the 202 response code. Each extension is listed on a separate line; the line MUST begin with an extension-label and optionally one or more parameters (separated by single spaces). The extension-label and the meaning of the parameters are specified as part of the definition of the extension. The extension-label is a string of 1 to 12 US-ASCII letters and MUST be in uppercase. Parameters are strings of 1 or more printable UTF-8 characters (that is, either printable US-ASCII characters or any UTF-8 sequence outside the US-ASCII range, but not space or TAB). The server MUST NOT list the same extension twice in the response, and MUST list all supported extensions. The order in which the extensions are listed is not significant. The server need not even consistently return the same order. If the server does not support any extensions, it MUST return an empty list. The 402 response code is documented for historic reasons only; clients SHOULD handle it gracefully, but servers MUST NOT generate it. Following a generic failure response, such as 403, an extension might still be available, and the client MAY attempt to use it. Example of a successful response: The particular extensions shown here are simply examples of what might be defined in other places, and no particular meaning should be attributed to them. Example where no extensions are available: Example from a non-conforming server which indicates "no extensions available" using the 402 response code: The client uses the QUIT command to terminate the session. The server MUST acknowledge the QUIT command and then close the connection to the client. This is the preferred method for a client to indicate that it has finished all its transactions with the NNTP server. If a client simply disconnects (or the connection times out or some other fault occurs), the server MUST gracefully cease its attempts to service the client, disconnecting from its end if necessary.

News reading clients have available a variety of mechanisms to retrieve articles via NNTP. The news articles are stored and indexed using three types of keys. One key is the message-id of an article. Another key is composed of the newsgroup name and the article number within that newsgroup. That key MUST be unique to a particular server (there will be only one article with that number within a particular newsgroup), but is not required to be globally unique. Additionally, because the same article can be cross-posted to multiple newsgroups, there may be multiple keys that point to the same article on the same server. The final key is the arrival timestamp, giving the time that the article arrived at the server. The server MUST ensure that article numbers are issued in order of arrival timestamp; that is, articles arriving later MUST have higher numbers than those that arrive earlier. The server SHOULD allocate the next sequential unused number to each new article. Article numbers MUST lie between 1 and 4,294,967,295 inclusive. The client and server SHOULD NOT use leading zeroes in specifying article numbers, and MUST NOT use more than 16 digits. In some situations, the value zero replaces an article number to show some special situation.

The following commands are used to set the "current selected newsgroup" and the "current article number", which are used by various commands. At the start of an NNTP session, both of these values are set to the special value "invalid". The required parameter is the name of the newsgroup to be selected (e.g. "news.software.b"). A list of valid newsgroups may be obtained by using the LIST ACTIVE command (see ). The successful selection response will return the article numbers of the first and last articles in the group at the moment of selection (these numbers are referred to as the "reported low water mark" and the "reported high water mark"), and an estimate of the number of articles on file in the group. If the group is not empty, the estimate MUST be at least the actual number of articles available, and MUST be no greater than one more than the difference between the reported low and high water marks. (Some implementations will actually count the number of articles on file. Others will just subtract the low water mark from the high water mark and add one to get an estimate.) If the group is empty, one of the following three situations will occur. Clients MUST accept all three cases; servers MUST NOT represent an empty group in any other way. The high water mark will be one less than the low water mark, and the estimated article count will be zero. Servers SHOULD use this method to show an empty group. This is the only time that the high water mark can be less than the low water mark. All three numbers will be zero. The high water mark is greater than or equal to the low water mark. The estimated article count might be zero or non-zero; if non-zero, the same requirements apply as for a non-empty group. The set of articles in a group may change after the GROUP command is carried out. That is: articles may be removed from the group articles may be reinstated in the group with the same article number, but those articles MUST have numbers no less than the reported low water mark (note that this is a reinstatement of the previous article, not a new article reusing the number) new articles may be added with article numbers greater than the reported high water mark (if an article that was the one with the highest number has been removed, the next new article will not have the number one greater than the reported high water mark) Except when the group is empty and all three numbers are zero, whenever a subsequent GROUP command for the same newsgroup is issued, either by the same client or a different client, the reported low water mark in the response MUST be no less than that in any previous response for that newsgroup sent to any client. The client may make use of the low water mark to remove all remembered information about articles with lower numbers, as these will never recur. This includes the situation when the high water mark is one less than the low water mark. No similar assumption can be made about the high water mark, as this can decrease if an article is removed, and then increase again if it is reinstated or if new articles arrive. When a valid group is selected by means of this command, the current selected newsgroup MUST be set to that group and the current article number MUST be set to the first article in the group. If an empty newsgroup is selected, the current article pointer is made invalid. If an invalid group is specified, the current selected newsgroup and current article number MUST NOT be changed. The GROUP command (or the LISTGROUP command, if implemented) MUST be used by a client and a successful response received before any other command is used that depends on the value of the current selected newsgroup or current article number. If the group specified is not available on the server, a 411 response MUST be returned. Example for a group known to the server: Example for a group unknown to the server: Example of an empty group using the preferred response: Example of an empty group using an alternative response: Example of an empty group using a different alternative response: If the current selected newsgroup is valid, the current article number MUST be set to the previous article in that newsgroup (that is, the highest existing article number less than the current article number). If successful, a response indicating the new current article number and the message-id of that article MUST be returned. No article text is sent in response to this command. There MAY be no previous article in the group, although the current article number is not the reported low water mark. There MUST NOT be a previous article when the current article number is the reported low water mark. Because articles can be removed and added, the results of multiple LAST and NEXT commands MAY not be consistent over the life of a particular NNTP session. If the current article number is already the first article of the newsgroup, a 422 response MUST be returned. If the current article number is invalid, a 420 response MUST be returned. If the current selected newsgroup is invalid, a 412 response MUST be returned. In all three cases the current selected newsgroup and current article number MUST NOT be altered. Example of a successful article retrieval using LAST: Example of an attempt to retrieve an article without having selected a group (via the GROUP command) first: Example of an attempt to retrieve an article using the LAST command when the current article number is that of the first article in the group: Example of an attempt to retrieve an article using the LAST command when the current selected newsgroup is empty: If the current selected newsgroup is valid, the current article number MUST be set to the next article in that newsgroup (that is, the lowest existing article number greater than the current article number). If successful, a response indicating the new current article number and the message-id of that article MUST be returned. No article text is sent in response to this command. If the current article number is already the last article of the newsgroup, a 421 response MUST be returned. In all other aspects (apart, of course, from the lack of 422 response) this command is identical to the LAST command. Example of a successful article retrieval using NEXT: Example of an attempt to retrieve an article without having selected a group (via the GROUP command) first: Example of an attempt to retrieve an article using the NEXT command when the current article number is that of the last article in the group: Example of an attempt to retrieve an article using the NEXT command when the current selected newsgroup is empty:

The ARTICLE, BODY, HEAD, and STAT commands are very similar. They differ only in the parts of the article that are presented to the client and in the successful response code. The ARTICLE command is described here in full, while the other commands are described in terms of the differences. As specified in , an article consists of two parts: the article headers and the article body. When responding to one of these commands, the server MUST present the entire article or appropriate part and MUST NOT attempt to alter or translate it in any way. The 420 response can only occur if no article number has been specified. The ARTICLE command selects an article based on the arguments and presents the entire article (that is, the headers, an empty line, and the body in that order). The command has two forms. In the first form, a message-id is specified (including the angle brackets), and the server presents the article with that message-id. In this case, the server MUST NOT alter the current selected newsgroup or current article number. This is both to facilitate the presentation of articles that may be referenced within another article being read, and because of the semantic difficulties of determining the proper sequence and membership of an article that may have been crossposted to more than one newsgroup. In the response, the article number is replaced with zero (that is, the server is not required to determine whether the article is in the current group or what article number(s) it has). In the second form, an article number may be specified. If so, and if there is an article with that number in the currently selected newsgroup, the server MUST set the current article number to that number. Then, whether or not a number was specified, the article indicated by the current article number is presented to the client. Note that a previously valid article number MAY become invalid if the article has been removed. A previously invalid article number MAY become valid if the article has been reinstated, but such an article number MUST be no less than the reported low water mark for that group. The server MUST NOT change the current selected newsgroup as a result of this command. The server MUST NOT change the current article number except when an article number argument was provided and the article exists; in particular, it MUST NOT change it following an unsuccessful response. Since the message-id is unique for each article, it may be used by a client to skip duplicate displays of articles that have been posted more than once, or to more than one newsgroup. The article is returned as a multi-line response following the 220 response code. If the current article number is invalid, a 420 response MUST be returned. If there is no article with the specified number, a 423 response MUST be returned. If the current selected newsgroup is invalid, a 412 response MUST be returned. Example of a successful retrieval of an article (using no article number): Example of a successful retrieval of an article by message-id: Example of an unsuccessful retrieval of an article by message-id: Example of an unsuccessful retrieval of an article by number: Example of an unsuccessful retrieval of an article by number because no newsgroup was selected first: Example of an attempt to retrieve an article when the current selected newsgroup is empty: The 420 response can only occur if no article number has been specified. The HEAD command behaves identically to the ARTICLE command except that, if the article exists, the response code is 221 instead of 220 and only the headers are presented (the empty line separating the headers and body MUST NOT be included). Example of a successful retrieval of the headers of an article (using no article number): Example of a successful retrieval of the headers of an article by message-id: Example of an unsuccessful retrieval of the headers of an article by message-id: Example of an unsuccessful retrieval of the headers of an article by number: Example of an unsuccessful retrieval the headers of an article by number because no newsgroup was selected first: Example of an attempt to retrieve the headers of an article when the current selected newsgroup is empty: The 420 response can only occur if no article number has been specified. The BODY command behaves identically to the ARTICLE command except that, if the article exists, the response code is 222 instead of 220 and only the body is presented (the empty line separating the headers and body MUST NOT be included). Example of a successful retrieval of the body of an article (using no article number): Example of a successful retrieval of the body of an article by message-id: Example of an unsuccessful retrieval of the body of an article by message-id: Example of an unsuccessful retrieval of the body of an article by number: Example of an unsuccessful retrieval of the body of an article by number because no newsgroup was selected first: Example of an attempt to retrieve the body of an article when the current selected newsgroup is empty: The 420 response can only occur if no article number has been specified. The STAT command behaves identically to the ARTICLE command except that, if the article exists, it is NOT presented to the client and the response code is 223 instead of 220. Note that the response is NOT multi-line. This command allows the client to determine whether an article exists, and in the second form what its message-id is, without having to process an arbitrary amount of text. Example of STAT on an existing article (using no article number): Example of a STAT of an existing article by message-id: Example of an STAT of an article not on the server by message-id: Example of STAT of an article not in the server by number: Example of STAT of an article by number when no newsgroup was selected first: Example of STAT of an article when the current selected newsgroup is empty:

Article posting is done in one of two modes: individual article posting from news reading clients using POST, and article transfer from other news servers using IHAVE. If posting is allowed, a 340 response MUST be returned to indicate that the article to be posted should be sent. If posting is prohibited for some installation-dependent reason, a 440 response MUST be returned. If posting is permitted, the article MUST be in the format specified in and MUST be sent by the client to the server in the manner specified in () for multi-line responses (except that there is no initial line containing a response code). Thus a single dot (".") on a line indicates the end of the text, and lines starting with a dot in the original text have that dot doubled during transmission. Following the presentation of the termination sequence by the client, the server MUST return a response indicating success or failure of the article transfer. Note that response codes 340 and 440 are used in direct response to the POST command. Others are returned following the sending of the article. A response of 240 SHOULD indicate that, barring unforseen server errors, the posted article will be made available on the server and/or transferred to other servers as appropriate. In other words, articles not wanted by the server SHOULD be rejected with a 441 response and not accepted and silently discarded. No attempt shall be made by the server to filter characters, fold or limit lines, or otherwise process incoming text. The intent is that the server just passes the incoming message to be posted to the server installation's news posting software, which is not defined by this document. The client SHOULD NOT assume that the article has been successfully transferred unless it receives an affirmative response from the server. If the session is interrupted before the response is received, it is possible that an affirmative response was sent but has been lost. Therefore, in any subsequent session, the client SHOULD either check whether the article was successfully posted before resending or, if the client supplied a message-id in the original article, ensure it supplies the same message-id - the latter approach is preferred since the article might not have been made available for reading yet (for example, it may have to go through a moderation process). In particular, if the article contained a header with name "Message-ID", the client SHOULD ensure that the contents of this header are identical when resending it and the server SHOULD ensure that the re-sent article is recognised as a duplicate and not assigned a different message-id to the original. Example of a successful posting: Example of an unsuccessful posting: Example of an attempt to post when posting is not allowed: The IHAVE command informs the server that the client has an article with the specified message-id. If the server desires a copy of that article a 335 response MUST be returned, instructing the client to send the entire article. If the server does not want the article (if, for example, the server already has a copy of it), a 435 response MUST be returned, indicating that the article is not wanted. Finally, if the article isn't wanted immediately but the client should retry later if possible (if, for example, another client is in the process of sending the same article to the server), a 436 response MUST be returned. If transmission of the article is requested, the client MUST send the entire article, including headers and body, in the format defined above () for multi-line responses (except that there is no initial line containing a response code). Thus a single dot (".") on a line indicates the end of the text, and lines starting with a dot in the original text have that dot doubled during transmission. The server MUST return either a 235 response, indicating that the article was successfully transferred, a 436 response, indicating that the transfer failed but should be tried again later, or a 437 response, indicating that the article was rejected. This function differs from the POST command in that it is intended for use in transferring already-posted articles between hosts. It SHOULD NOT be used when the client is a personal news reading program, since use of this command indicates that the article has already been posted at another site and is simply being forwarded from another host. However, despite this, the server MAY elect not to post or forward the article if, after further examination of the article, it deems it inappropriate to do so. Reasons for such subsequent rejection of an article may include such problems as inappropriate newsgroups or distributions, disc space limitations, article lengths, garbled headers, and the like. These are typically restrictions enforced by the server host's news software and not necessarily the NNTP server itself. The client SHOULD NOT assume that the article has been successfully transferred unless it receives an affirmative response from the server. A lack of response (such as a dropped network connection or a network timeout) SHOULD be treated the same as a 436 response. Because some news server software may not be able immediately to determine whether or not an article is suitable for posting or forwarding, an NNTP server MAY acknowledge the successful transfer of the article (with a 235 response) but later silently discard it. Example of successfully sending an article to another site: Example of sending an article to another site that rejects it: Example of sending an article to another site where the transfer fails: Example of sending an article to a site that already has it: Example of sending an article to a site that requests the article be tried again later:

This section lists other commands that may be used at any time between the beginning of a session and its termination. Using these commands does not alter any state information, but the response generated from their use may provide useful information to clients. This command exists to help clients find out the current Coordinated Universal Time from the server's perspective. This command SHOULD NOT be used as a substitute for NTP but to provide information that might be useful when using the NEWNEWS command (see ). A system providing NNTP service SHOULD keep the system clock as accurate as possible, either with NTP or by some other method. The server MUST return a 111 response specifying the date and time on the server in the form yyyymmddhhmmss. This date and time is in Coordinated Universal Time. This command provides a short summary of commands that are understood by this implementation of the server. The help text will be presented as a multiline response following the 100 response code. This text is not guaranteed to be in any particular format and MUST NOT be used by clients as a replacement for the LIST EXTENSIONS command described in This command returns a list of newsgroups created on the server since the specified date and time. The results are in the same format as the LIST ACTIVE command (see ). However, they MAY include groups not available on the server (and so not returned by LIST ACTIVE) and MAY omit groups for which the creation date is not available. The results SHOULD be consistent with those of the LIST ACTIVE.TIMES command, except that if the specified date and time is earlier than the oldest entry in the latter then the results of this command may include extra groups. The date is specified as 6 or 8 digits in the format [xx]yymmdd, where xx is the first two digits of the year (19-99), yy is the last two digits of the year (00-99), mm is the month (01-12), and dd is the day of the month (01-31). Clients SHOULD specify all four digits of the year. If the first two digits of the year are not specified (this is supported only for backwards compatibility), the year is to be taken from the current century if yy is smaller than or equal to the current year, otherwise the year is from the previous century. The time is specified as 6 digits in the format hhmmss, where hh is the hours in the 24-hour clock (00-23), mm is the minutes (00-59), and ss is the seconds (00-60, to allow for leap seconds). The token "GMT" specifies that the date and time are given in Coordinated Universal Time; if it is omitted then the date and time are specified in the server's local timezone. Note that there is no way using the protocol specified in this document to establish the server's local timezone. Note that an empty list is a possible valid response and indicates that there are no new newsgroups since that date-time. Clients SHOULD make all queries using Coordinated Universal Time (i.e. by including the "GMT" parameter) when possible. Example where there are new groups: Example where there are no new groups: This command returns a list of message-ids of articles posted or received on the server, in the newsgroups whose names match the wildmat, since the specified date and time. One message-id is sent on each line; the order of the response has no specific significance and may vary from response to response in the same session. A message-id MAY appear more than once; if it does so, it has the same meaning as if it appeared only once. Date and time are in the same format as the NEWGROUPS command (see ). Note that an empty list is a possible valid response and indicates that there is currently no new news in the relevant groups. Clients SHOULD make all queries in Coordinated Universal Time (i.e. by using the "GMT" parameter) when possible. Example where there are new articles: Example where there are no new articles:

As described in , each article has an arrival timestamp. Each newsgroup also has a creation timestamp. These timestamps are used by the NEWNEWS and NEWGROUP commands to construct their reponses. The DATE command MUST return a timestamp from the same clock as is used for determining article arrival and group creation times. This clock SHOULD be monotonic, and adjustments SHOULD be made by running it fast or slow compared to "real" time rather than by making sudden jumps. Clients can ensure that they do not have gaps in lists of articles or groups by using the DATE command in the following manner: Issue DATE command and record result Issue NEWNEWS command using a previously chosen timestamp Issue DATE command and hold result in temporary storage Issue NEWNEWS command using timestamp saved from previous session Overwrite saved timestamp with that currently in temporary storage In order to allow for minor errors, clients MAY want to adjust the timestamp back by two or three minutes before using it in NEWNEWS. First session: Second session (the client has subtracted 3 minutes from the timestamp returned previously): Note how <article.3@local.service> arrived in the 3 minute gap and so is listed in both responses.

The LIST ACTIVE command with no parameters returns a list of valid newsgroups and associated information. The server MUST include every group that the client is permitted to select with the GROUP command. Each newsgroup is sent as a line of text in the following format: where: is the name of the newsgroup; is the reported high water mark for the group; is the reported low water mark for the group; is the current status of the group on this server. Each field in the line is separated from its neighboring fields by one or more spaces. Note that an empty list is a possible valid response, and indicates that there are currently no valid newsgroups. The reported high and low water marks are as described in the GROUP command (see ). The status field is typically one of: posting is permitted posting is not permitted postings will be forwarded to the newsgroup moderator The server SHOULD use these values when these meanings are required and MUST NOT use them with any other meaning. Other values for the status may exist; the definition of these other values and the circumstances under which they are returned may be specified in an extension or may be private to the server. A client SHOULD treat an unrecognised status as giving no information. The status of a newsgroup only indicates how posts to that newsgroup are normally processed and is not necessarily customised to the specific client. For example, if the current client is forbidden from posting, then this will apply equally to groups with status "y". Conversely, a client with special privileges (not defined by this specification) might be able to post to a group with status "n". If the optional wildmat parameter is specified, the list is limited to only the groups whose names match the wildmat. If no wildmat is specified, the keyword ACTIVE MAY be omitted without altering the effect of the command. Example of LIST ACTIVE returning a list of newsgroups: Example of LIST ACTIVE omitting the second keyword and returning no newsgroups: Example of LIST ACTIVE with a wildmat: The active.times file is maintained by some news transport systems to contain information about who created a particular newsgroup and when. Each line of this file consists of three fields separated from each other by one or more spaces. The first field is the name of the newsgroup. The second is the time when this group was created on this news server, measured in seconds since the start of January 1, 1970. The third is the email address of the entity that created the newsgroup, and must be a mailbox as defined in RFC 2822. Should the third field simply be free-form, or should it be recommended usage rather than mandatory? The problem with "mailbox" is that mailbox requires that it be fully qualified, and unqualified addresses are apparently very common for groups created directly by the administrator. The file MAY omit newsgroups for which the information is unavailable and MAY include groups not available on the server; in particular, the file MAY omit all groups created before the date and time of the oldest entry. The client MUST NOT assume that the list is complete or that it matches the list returned by LIST ACTIVE. The NEWGROUPS command may provide a better way to access this information and the results of the two commands SHOULD be consistent (subject to the caveats in the description of that command). If the information is available, it is returned as a multi-line response following the 215 response code. If the optional wildmat parameter is specified, the list is limited to only the groups in the file whose names match the wildmat. Note that an empty list is a possible valid response, and indicates that there are no groups in the file, or that match the wildmat. Example of LIST ACTIVE.TIMES returning a list of newsgroups: Example of LIST ACTIVE.TIMES returning an error where the command is recognised but the software does not maintain this information: Example of LIST ACTIVE.TIMES sent to a server that does not recognize this command: The distributions file is maintained by some news transport systems to contain information about valid values for the content of the Distribution header in a news article and about what the various values mean. Each line of this file consists of two fields separated from each other by one or more spaces. The first field is a value and the second is a short explanation of the meaning of that value. If the information is available, it is returned as a multi-line response following the 215 response code. Example of LIST DISTRIBUTIONS returning a list of distributions: Example of LIST DISTRIBUTIONS returning an error where the command is recognised but the software does not maintain this information: Example of LIST DISTRIBUTIONS sent to a server that does not recognize this command: The distrib.pats file is maintained by some news transport systems to choose a value for the content of the Distribution header of a news article being posted. Each line of this file consists of three fields separated from each other by a colon (":"). The first field is a weight, the second field is a wildmat (which may be a simple group name), and the third field is a value for the Distribution header content. The client MAY use this information to construct an appropriate Distribution header given the name of a newsgroup. To do so, it should determine the lines whose second field matches the newsgroup name, select from among them the line with the highest weight (with 0 being the lowest), and use the value of the third field to construct the Distribution header. If the information is available, it is returned as a multi-line response following the 215 response code. Example of LIST DISTRIB.PATS returning a list of newsgroups: Example of LIST DISTRIB.PATS returning an error where the command is recognised but the software does not maintain this information: Example of LIST DISTRIB.PATS sent to a server that does not recognize this command: The newsgroups file is maintained by some news transport systems to contain the name of each newsgroup that is available on the server and a short description about the purpose of the group. Each line of this file consists of two fields separated from each other by one or more space or TAB characters (usual practice is a single TAB). The first field is the name of the newsgroup and the second is a short description of the group. Note that an empty list is a possible valid response, and indicates that there are currently no valid newsgroups. The file MAY omit newsgroups for which the information is unavailable and MAY include groups not available on the server. The client MUST NOT assume that the list is complete or that it matches the list returned by LIST ACTIVE. If the information is available, it is returned as a multi-line response following the 215 response code. If the optional wildmat parameter is specified, the list is limited to only the groups in the file whose names match the wildmat. Note that an empty list is a possible valid response, and indicates that there are no groups in the file, or that match the wildmat. Example of LIST NEWSGROUPS returning a list of newsgroups: Example of LIST NEWSGROUPS returning an error where the command is recognised but the software does not maintain this information: Example of LIST NEWSGROUPS sent to a server that does not recognize this command:

Although NNTP is widely and robustly deployed, some parts of the Internet community might wish to extend the NNTP service. This document defines a means whereby an extended NNTP client can query the server to determine the service extensions that it supports. It must be emphasized that any extension to the NNTP service should not be considered lightly. NNTP's strength comes primarily from its simplicity. Experience with many protocols has shown that: Protocols with few options tend towards ubiquity, whilst protocols with many options tend towards obscurity. This means that each and every extension, regardless of its benefits, must be carefully scrutinized with respect to its implementation, deployment, and interoperability costs. In many cases, the cost of extending the NNTP service will likely outweigh the benefit. Given this environment, the framework for extensions described in this document consists of: a mechanism for clients to determine a server's available extensions a registry of NNTP service extensions The LIST EXTENSIONS command is described in this document (see ) and is the mechanism for clients to use to determine what extensions are available. The IANA shall maintain a registry of NNTP service extensions. An extension is identified by a unique extension-label, which is a string of 1 to 12 uppercase US-ASCII letters. The extension-label will often be the name of a new command that the extension adds. However this is not a requirement: an extension might not add any new commands or keywords. An extension is either a private extension or else it is included in the IANA registry and is defined in an RFC. Such RFCs either must be on the standards-track or must define an IESG-approved experimental protocol. The definition of an extension must include: a descriptive name for the extension the extension-label (which is returned by LIST EXTENSIONS to indicate to the client that the server supports this particular extension) the syntax, values, and meanings of any parameters following the extension-label in the output of LIST EXTENSIONS any new NNTP commands associated with the extension the syntax and possible values of parameters associated with the new NNTP commands the response codes and possible values of parameters for the responses of the new NNTP commands any new parameters the extension associates with any other pre-existing NNTP commands how support for the extension affects the behavior of a server and NNTP client any increase in the maximum length of commands over the value specified in this document a specific statement about the effect on pipelining this extension may have (if any) The extension-label of private extensions MUST begin with "X". The extension-label of registered extensions MUST NOT begin with "X". A server MUST NOT provide any extension, whether or not listed in the output from LIST EXTENSIONS, unless it is either a registered extension or a private extension. As worded, this forbids commands like MODE SLAVE that servers already provide but that aren't part of an existing extension. We can't simply make these illegal. The wording about starting keywords with an X could be reduced to a SHOULD, except for backwards compatibility (with a pointer to RFC 2980). But is that the right answer? Except where stated otherwise, the commands in this document are understood (even if not supported) by all servers and are not described in the list of features returned by the LIST EXTENSIONS command. A server MAY provide additional keywords - either for new commands or new variants of existing commands - as part of a private extension. These new keywords MUST begin with "X". A server MUST NOT send different response codes to basic NNTP commands documented here or commands documented in registered extensions in response to the availability or use of a private extension.

The IANA's initial registry of NNTP service extensions consists of these entries: Extension Label Added behavior Specific article numbers LISTGROUP Defined in this document Overview support OVER Defined in this document Header pattern matching HDR Defined in this document

Each of the following sections describes an extension that a server MAY provide. If the server provides the extension, it MUST include the appropriate extension label in the response to LIST EXTENSIONS. If it does not provide it, it MUST NOT include the appropriate extension label. The descriptions of facilities in each section are written as if the extension is provided. If it is not provided, the entire section should be ignored. If the server provides an extension, it MUST implement all of the commands in the specification of the extension except for those marked as optional. If it does not provide an extension, it MUST NOT implement any of the commands in the specification of that extension.

This extension provides one command and has the extension label LISTGROUP. The 412 response can only occur if no group has been specified. The LISTGROUP command is used to get a listing of all the article numbers in a particular newsgroup. The optional parameter is the name of the newsgroup to be selected (e.g. "news.software.misc"). A list of valid newsgroups may be obtained from the LIST ACTIVE command. If no group is specified, the current selected newsgroup is used. The list of article numbers is returned as a multi-line response following the 211 response code (the parameters on the initial response line are the same as for the GROUP command (see ). The list contains one number per line, is in numerical order, and lists precisely those articles that exist in the group. When a valid group is selected by means of this command, the current selected newsgroup MUST be set to that group and the current article number MUST be set to the first article in the group. If an empty newsgroup is selected, the current article pointer is made invalid. If an invalid group is specified, the current selected newsgroup and current article number MUST NOT be changed. The LISTGROUP command MAY be used by a client as a replacement for the GROUP command in establishing a valid current selected newsgroup and current article number. If the group specified is not available on the server, a 411 response MUST be returned. If no group is specified and the current selected newsgroup is invalid, a 412 response MUST be returned. Example of LISTGROUP on an empty group: Example of LISTGROUP on a valid current selected newsgroup: Example of LISTGROUP failing because no group has been selected:

The OVER and HDR extensions refer to the concept of "article metadata". This is data about articles that does not occur within the article itself. Each metadata item has a name which MUST begin with a colon (and which MUST NOT contain a colon elsewhere within it). When generating a metadata item, the server MUST compute it for itself and MUST NOT trust any related value provided in the article. (In particular, a Lines or Bytes header in the article MUST NOT be assumed to specify the correct number of lines or bytes in the article.) This specification defines two metadata items: ":bytes" and ":lines". Implementations and other extensions may define other metadata items. Do we need a separate private namespace? For example, we could reserve :name for extensions and ::name for implementation use.

The :bytes metadata item for an article is a decimal integer. It MUST equal the number of octets in the entire article - headers, body, and separating empty line - except that each CRLF pair MAY (but SHOULD NOT) be counted as a single octet. Should this be called ":octets" instead?

The :lines metadata item for an article is a decimal integer. It MUST equal the number of lines in the article body (excluding the empty line separating headers and body); equivalently, it is two less than the number of CRLF pairs that the BODY command would return for that article (the extra two are those following the response code and the termination octet).

This extension provides two commands, OVER and LIST OVERVIEW.FMT. The label for this extension is OVER. The OVER extension provides access to the "overview database", which is a database of headers extracted from incoming articles. Only certain headers are included in the database. The database also includes some article metadata. The information stored in the database may change over time. The LIST OVERVIEW.FMT command describes the information that would be stored for an article arriving at the same time as the command was executed. This extension is based on the Overview/NOV database developed by Geoff Collyer. The OVER command returns the contents of the headers and metadata in the database for the article(s) specified from the current selected newsgroup. The optional range argument may be any of the following: an article number an article number followed by a dash to indicate all following an article number followed by a dash followed by another article number If no argument is specified, then the current article number is used. If the information is available, it is returned as a multi-line response following the 224 response code. If the current selected newsgroup is invalid, a 412 response MUST be returned. If there are no articles in the range specified, a 423 response MUST be returned. If OVER is sent without any arguments and the current article number is invalid, a 420 response MUST be returned. For a successful response, the output consists of one line per article, sorted in numerical order of article number. Each line consists of a number of fields separated by a TAB. A field may be empty (in which case there will be two adjacent TABs), and a sequence of trailing TABs may be omitted. The first 8 fields MUST be the following, in order: article number Subject header content From header content Date header content Message-ID header content References header content :bytes metadata item :lines metadata item Any subsequent fields are the contents of the other headers and metadata held in the database. For the five mandatory headers, the content of each field MUST be based on the content of the header (that is, with the header name and following colon and space removed). If the article does not contain that header, or if the content is empty, the field MUST be empty. For the two mandatory metadata items, the content of the field MUST be just the value, with no other text. For all subsequent fields that contain headers, the content MUST be the entire header line other than the trailing CRLF. For all subsequent fields that contain metadata, the field consists of the metadata name, a single space, and then the value. For all fields, the value is processed by first removing all CRLF pairs (that is, undoing any folding and removing the terminating CRLF) and then replacing each TAB with a single space. If there is no such header in the article, or no such metadata item, or no header or item stored in the database for that article, the corresponding field MUST be empty. Note that, after unfolding, the characters NUL, LF, and CR cannot occur in the header of an article offered by a conformant server. Nevertheless, servers SHOULD check for these characters and replace each one by a single space (so that, for example, CR LF LF TAB will become two spaces, since the CR and first LF will be removed by the unfolding process). This will encourage robustness in the face of non-conforming data; it is also possible that future versions of this specification may permit these characters to appear in articles. The server SHOULD NOT produce output for articles that no longer exist. In the first two examples, TAB has been replaced by vertical bar and some lines have been folded for readability. Example of a successful retrieval of overview information for an article (using no article number): Example of a successful retrieval of overview information for a range of articles: Note the missing "References" and Xref headers in the second line, the missing trailing field(s) in the first and last lines, and that there are only results for those articles that still exist. Example of an unsuccessful retrieval of overview information on an article by number: Example of an unsuccessful retrieval of overview information by number because no newsgroup was selected first: Example of an attempt to retrieve information when the current selected newsgroup is empty: Should this be optional even when the OVER extension is provided? Or even just removed entirely? What do we want to require about the OVER contents being consistent with the output of this command? The LIST OVERVIEW.FMT command returns a description of the fields in the database. The fields MUST be listed in the order that they will be returned by the OVER command for a newly-received article (the information stored for articles may change over time). If the information is available, it is returned as a multi-line response following the 215 response code. The information contains one line per field in the order they are returned by the OVER command; the first 7 lines MUST be exactly:

Subject: From: Date: Message-ID: References: :bytes :lines except that, for compatibility with existing implementations, the last two lines MAY instead be:

Bytes: Lines: even though they refer to metadata, not headers. All subsequent lines MUST consist of either a header name followed by ":full", or the name of a piece of metadata. There are no leading or trailing spaces in the output. Note that the 7 fixed lines describe the 2nd to 8th fields of the OVER output. The "full" suffix is a reminder that the corresponding fields include the header name. This command MAY generate different results if used more than once in a session. Example of LIST OVERVIEW.FMT output corresponding to the example OVER output above, using the preferred format: Example of LIST OVERVIEW.FMT output corresponding to the example OVER output above, using the alternative format: Example of LIST OVERVIEW.FMT returning an error:

This extension provides one new command: HDR. The label for this extension is HDR. There is ongoing discussion about whether this extension should have a parameter and, if so, what it means. The HDR command retrieves specific headers from an article or specified range of articles in the current selected newsgroup, or from an article specified by message-id. It can also return certain metadata about the article or articles. The required header parameter is the name of a header (e.g. "subject") in an article, or the name of a metadata item, and is case-insensitive. Names of metadata items always begin with a colon. Except where stated otherwise, metadata items are treated as if they were header contents, and references to headers in this description apply equally to metadata items. The range parameter may be any of the following: an article number an article number followed by a dash to indicate all following an article number followed by a dash followed by another article number The message-id argument indicates a specific article. As shown by the syntax, the range and message-id arguments are mutually exclusive; if neither is specified, the current article number is used. If the information is available, it is returned as a multi-line response following the 225 response code and contains one line for each article where the relevant header line exists. The line consists of the article number, a space, and then the contents of the header (without the header name or the colon and space that follow it) or metadata item. If the article is specified by message-id rather than by article range, the article number is given as "0". Header contents are modified as follows: all CRLF pairs are removed, and then each TAB is replaced with a single space (note that this is the same transformation as is performed by the OVER extension, and the same comment concerning NUL, CR, and LF applies). The header content is in all cases taken from the article. This means that, for example, a request for the header "Lines" returns the contents of the "Lines" header of the specified articles, if any, not the line count metadata or any other server-generated value. If the header occurs in a given article multiple times, only the content of the first occurrence is returned by HDR. If the requested header is not present in the article or if it is present but empty, a line for that article is included in the output but the header content portion of the line is empty (the space after the article number MAY be retained or omitted). If any article number in the provided range does not exist in the group, no line for that article number is included in the output. If the optional argument is a message-id and no such article exists, a 430 response MUST be returned. If the optional argument is not a message-id and the current selected newsgroup is invalid, a 412 response MUST be returned. If the optional argument is an article number or number range and no article with that number or in that number range exists in the current selected newsgroup, a 423 response MUST be returned. If HDR is sent without any arguments and the current article number is invalid, a 420 response MUST be returned. A server MAY only allow HDR commands for a limited set of headers and metadata items (such as those present in the overview database). If so, it MUST respond with a 503 response to attempts to request other headers, rather than returning erroneous results such as a successful empty response. Example of a successful retrieval of subject lines from a range of articles (3000235 has no Subject header, and 3000236 is missing): Example of a successful retrieval of line counts from a range of articles: Example of a successful retrieval of the subject line from an article by message-id: Example of a successful retrieval of the subject line from the current article: Example of an unsuccessful retrieval of a header from an article by message-id: Example of an unsuccessful retrieval of headers from articles by number because no newsgroup was selected first: Example of an unsuccessful retrieval of headers because the current selected newsgroup is empty: Example of an unsuccessful retrieval of headers because the server does not allow HDR commands for that header:

Each of the following sections describes the syntax of a major element of NNTP. This syntax extends and refines the descriptions elsewhere in this specification, and should be given precedence when resolving apparent conflicts. Note that ABNF strings are case insensitive. Non-terminals used in several places are defined in a separate section at the end.

This syntax defines the non-terminal "command-line", which represents what is sent from the client to the server. command-line = command EOL command = article-command / body-command / date-command / group-command / hdr-command / head-command / help-command / ihave-command / last-command / list-active-command / list-active-times-command / list-distrib-pats-command / list-distributions-command / list-extensions-command / list-newsgroups-command / list-overview-fmt-command / listgroup-command / mode-reader-command / newgroups-command / newnews-command / next-command / over-command / post-command / quit-command / stat-command / x-command article-command = "ARTICLE" [article-ref] body-command = "BODY" [article-ref] date-command = "DATE" group-command = "GROUP" WS newsgroup-name hdr-command = "HDR" WS header-meta-name [range-ref] head-command = "HEAD" [article-ref] help-command = "HELP" ihave-command = "IHAVE" WS message-id last-command = "LAST" list-active-command = "LIST" [WS "ACTIVE" [WS wildmat]] list-active-times-command = "LIST" WS "ACTIVE.TIMES" [WS wildmat] list-distrib-pats-command = "LIST" WS "DISTRIB.PATS" list-distributions-command = "LIST" WS "DISTRIBUTIONS" list-extensions-command = "LIST" WS "EXTENSIONS" list-newsgroups-command = "LIST" WS "NEWSGROUPS" [WS wildmat] list-overview-fmt-command = "LIST" WS "OVERVIEW.FMT" listgroup-command = "LISTGROUP" [WS newsgroup-name] mode-reader-command = "MODE" WS "READER" newgroups-command = "NEWGROUPS" WS date-time newnews-command = "NEWNEWS" WS wildmat WS date-time next-command = "NEXT" over-command = "OVER" [WS range] post-command = "POST" quit-command = "QUIT" stat-command = "STAT" [article-ref] x-command = x-command-name *(WS x-argument) ; Each extension command is specified fully elsewhere article-ref = WS (article-number / message-id) article-number = 1*16DIGIT date = [2DIGIT] 6DIGIT date-time = date WS time [WS "GMT"] header-meta-name = header-name / metadata-name metadata-name = ":" 1*A-NOTCOLON newsgroup-name = 1*wildmat-exact range = article-number ["-" [article-number]] range-ref = WS (range / message-id) time = 6DIGIT x-command-name = 3*12A-CHAR x-argument = 1*P-CHAR wildmat = wildmat-pattern *("," ["!"] wildmat-pattern) wildmat-pattern = 1*wildmat-item wildmat-item = wildmat-exact / wildmat-wild wildmat-exact = %x21-29 / %x2B / %x2D-3E / %x40-5A / %x5E-7E / UTF8-non-ascii ; exclude * , ? [ \ ] wildmat-wild = "*" / "?"

This syntax defines the non-terminal "response", which represents what is sent from the server to the client in response to a command. response = simple-response / multiline-response multiline-response = simple-response *content-line termination termination = "." CRLF content-line = [content-text] CRLF content-text = (".." / B-NONDOT) B-CHAR simple-response = 3DIGIT parameters [ SP trailing-comment ] CRLF trailing-comment = *U-CHAR parameters = *( SP parameter ) ; How many depends on the response parameter = 1*A-CHAR

This syntax defines the non-terminal "article", which represents the format of an article as described in . article = 1*header CRLF body body = *(*B-CHAR CRLF) header = header-name ":" header-tail CRLF header-tail = SP header-content-u / CRLF SP header-content-f header-content-u = *( header-gap header-text) *WS header-content-f = *WS header-text header-content-u header-gap = *WS [CRLF] 1*WS header-text = 1*P-CHAR

header-name = 1*A-NOTCOLON message-id = "<" 1*248A-NOTGT ">" ; Assorted special character sets ; A- means based on ASCII, excluding controls and SP ; P- means based on UTF-8, excluding controls and SP ; U- means based on UTF-8, excluding NUL CR and LF ; B- means based on bytes, excluding NUL CR and LF A-CHAR = %x21-7E A-NOTCOLON = %x21-39 / %x3B-7E ; exclude ":" A-NOTGT = %x21-3D / %x3F-7E ; exclude ">" P-CHAR = A-CHAR / UTF8-non-ascii U-CHAR = %x01-09 / %x0B-0C / %x0E-7F / UTF8-non-ascii B-CHAR = %x01-09 / %x0B-0C / %x0E-FF B-NONDOT = %x01-09 / %x0B-0C / %x0E-2D / %x2F-FF ; exclude "." CR = %x0D CRLF = CR LF DIGIT = %x30-39 EOL = *(SP / HT) CRLF HT = %x09 LF = %x0A SP = %x20 UTF8-non-ascii = UTF8-2 / UTF8-3 / UTF8-4 UTF8-2 = %xC2-DF UTF8-tail UTF8-3 = %xE0 %xA0-BF UTF8-tail / %xE1-EC 2UTF8-tail / %xED %x80-9F UTF8-tail / %xEE-EF 2UTF8-tail UTF8-4 = %xF0 %x90-BF 2UTF8-tail / %xF1-F3 3UTF8-tail / %xF4 %x80-8F 2UTF8-tail UTF8-tail = %x80-BF WS = 1*(SP / HT) When draft-yergeau-rfc2279bis-04.txt replaces 2279, need to update references.

This specification requires IANA to keep a registry of extension-labels. The initial contents of this registry are specified in . As described in , names beginning with X are reserved for private use while all other names are to be associated with a specification in an RFC on the standards-track or defining an IESG-approved experimental protocol.

This section is meant to inform application developers, information providers, and users of the security limitations in NNTP as described by this document. The discussion does not include definitive solutions to the problems revealed, though it does make some suggestions for reducing security risks.

NNTP, because it was created to distribute network news articles, will forward whatever information is stored in those articles. Specification of that information is outside this scope of this document, but it is likely that some personal and/or proprietary information is available in some of those articles. It is very important that designers and implementers provide informative warnings to users so personal and/or proprietary information in material that is added automatically to articles (e.g. in headers) is not disclosed inadvertently. Additionally, effective and easily understood mechanisms to manage the distribution of news articles SHOULD be provided to NNTP Server administrators, so that they are able to report with confidence the likely spread of any particular set of news articles.

A server is in the position to save session data about a user's requests that might identify their reading patterns or subjects of interest. This information is clearly confidential in nature and its handling can be constrained by law in certain countries. People using the NNTP protocol to provide data are responsible for ensuring that such material is not distributed without the permission of any individuals that are identifiable by the published results.

There is no user-based or token-based authentication in the basic NNTP specification. Access is normally controlled by server configuration files. Those files specify access by using domain names or IP addresses. However, this specification does permit the creation of extensions to the NNTP protocol itself for such purposes. While including such mechanisms is optional, doing so is strongly encouraged. Other mechanisms are also available. For example, a proxy server could be put in place that requires authentication before connecting via the proxy to the NNTP server.

Many existing NNTP implementations authorize incoming connections by checking the IP address of that connection against the IP addresses obtained via DNS lookups of lists of domain names given in local configuration files. Servers that use this type of authentication, and clients that find a server by doing a DNS lookup of the server name, rely very heavily on the Domain Name Service, and are thus generally prone to security attacks based on the deliberate misassociation of IP addresses and DNS names. Clients and servers need to be cautious in assuming the continuing validity of an IP number/DNS name association. In particular, NNTP clients and servers SHOULD rely on their name resolver for confirmation of an IP number/DNS name association, rather than caching the result of previous host name lookups. Many platforms already can cache host name lookups locally when appropriate, and they SHOULD be configured to do so. It is proper for these lookups to be cached, however, only when the TTL (Time To Live) information reported by the name server makes it likely that the cached information will remain useful. If NNTP clients or servers cache the results of host name lookups in order to achieve a performance improvement, they MUST observe the TTL information reported by DNS. If NNTP clients or servers do not observe this rule, they could be spoofed when a previously accessed server's IP address changes. As network renumbering is expected to become increasingly common, the possibility of this form of attack will grow. Observing this requirement thus reduces this potential security vulnerability. This requirement also improves the load-balancing behavior of clients for replicated servers using the same DNS name and reduces the likelihood of a user's experiencing failure in accessing sites that use that strategy.

UTF-8 permits only certain sequences of octets and designates others as either malformed or "illegal". The Unicode standard identifies a number of security issues related to illegal sequences and forbids their generation by conforming implementations. Implementations of this specification MUST NOT generate malformed or illegal sequences and SHOULD detect them and take some appropriate action. This could include: replacing such sequences by a "guessed" valid sequence (based on properties of the UTF-8 encoding); replacing such sequences by the sequence %xEF.BF.BD, which encodes the "replacement character" U+FFFD; closing the connection; generating a 501 response code. In the first case, the implementation MUST ensure that any replacement cannot be used to bypass validity or security checks. For example, the illegal sequence %xC0.A0 is an over-long encoding for space (%x20). If it is replaced by the latter in a command line, this needs to happen before the command line is parsed into individual arguments. If the replacement came after parsing, it would be possible to generate an argument with an embedded space, which is forbidden. Use of the "replacement character" does not have this problem, since it is permitted wherever non-US-ASCII characters are. Yergeau says that you MUST detect illegal sequences. He also rejects the first bullet point and consequent text; I'm discussing it with him now.

The author acknowledges the original authors of NNTP as documented in RFC 977: Brian Kantor and Phil Lapsey. The author gratefully acknowledges the work of the NNTP committee chaired by Eliot Lear. The organization of this document was influenced by the last available draft from this working group. A special thanks to Eliot for generously providing the original machine-readable sources for that document. The author gratefully acknowledges the work of Marshall Rose & John G. Meyers in RFC 1939 and the work of the DRUMS working group, specifically RFC 1869, which is the basis of the NNTP extensions mechanism detailed in this document. Why RFC 1939? The author gratefully acknowledges the authors of RFC 2616 for providing specific and relevant examples of security issues that should be considered for HTTP. Since many of the same considerations exist for NNTP, those examples that are relevant have been included here with some minor rewrites. The author gratefully acknowledges the comments and additional information provided by the following individuals in preparing one or more of the progenitors of this document: Russ Allbery <rra@stanford.edu> Wayne Davison <davison@armory.com> Chris Lewis <clewis@bnr.ca> Tom Limoncelli <tal@mars.superlink.net> Eric Schnoebelen <eric@egsner.cirr.com> Rich Salz <rsalz@osf.org> This work was motivated by the work of various news reader authors and news server authors, which includes those listed below: Original author of the NNTP extensions to the RN news reader and last maintainer of Bnews Original author of the NNTP extensions to the news readers that are part of Bnews Original author of the OVERVIEW database proposal and one of the original authors of CNEWS Original author of the xvnews news reader Author of the first threading extensions to the RN news reader (commonly called TRN) Original author of ANU NEWS Original author of the UNIX reference implementation for NNTP Original maintainer of the TIN news reader First known implementer of the AUTHINFO GENERIC extension Original author of INN One of the original authors of CNEWS Original author of the NN news reader Finally, the present author gratefully acknowledges the vast amount of work put into previous drafts by the previous author: Stan Barber <sob@academ.com>