NAME

SYNOPSIS

DESCRIPTION

Unlike innd(8), nnrpd supports all NNTP commands for user-oriented reading and posting. nnrpd uses the readers.conf file to control who is authorized to access the Usenet database.

On exit, nnrpd will report usage statistics through syslog(3).

nnrpd only reads config files (both readers.conf and inn.conf) when it is spawned. You can therefore never change the behavior of a client that's already connected. If nnrpd is run from innd (the default) or from inetd(8), xinetd(8), or some equivalent, a new nnrpd process is spawned for every connection and therefore any changes to configuration files will be immediately effective for all new connections. If you are instead running nnrpd with the -D option, any configuration changes won't take effect until nnrpd is restarted.

The inn.conf setting nnrpdflags can be used to pass any of the options below to instances of nnrpd that are spawned directly from innd. Many options only make sense when -D is used, so these options should not be used with nnrpdflags. See also the discussion of nnrpdflags in inn.conf(5).

When nnrpdloadlimit in inn.conf is not 0, it will also reject connections if the load average is greater than that value (typically 16). nnrpd can also prevent high-volume posters from abusing your resources. See the discussion of exponential backoff in inn.conf(5).

nnrpd injects articles into the local server running innd through a UNIX domain socket, or an INET domain socket if not supported. If another server should be used for injection, you can set it with the nnrpdposthost parameter in inn.conf. In case authentication credentials are requested during the injection, nnrpd will use the passwd.nntp file in pathetc.

OPTIONS

-4 address

The -4 parameter instructs nnrpd to bind to the specified IPv4 address when started as a standalone daemon using the -D flag. This has to be a valid IPv4 address belonging to an interface of the local host. It can also be 0.0.0.0, saying to bind to all addresses (this is the default).

-6 address

The -6 parameter instructs nnrpd to bind to the specified IPv6 address when started as a standalone daemon using the -D flag. This has to be a valid IPv6 address belonging to an interface of the local host. It can also be "::0", saying to bind to all IPv6 addresses.

By default, nnrpd in daemon mode listens to both IPv4 and IPv6 addresses. With this option, it will listen only to the specified IPv6 addresses. On some systems however, a value of "::0" will cause it to listen to all IPv4 addresses as well.

-b address

Similar to the -4 flag. -b is kept for backwards compatibility.

-c configfile

By default, nnrpd reads the readers.conf to determine how to authenticate connections. The -c flag specifies an alternate file for this purpose. If the file name isn't fully qualified, it is taken to be relative to pathetc in inn.conf. (This is useful to have several instances of nnrpd running on different ports or IP addresses with different settings.)

-D

If specified, this parameter causes nnrpd to operate as a daemon. That is, it detaches itself and runs in the background, forking a process for every connection. By default, nnrpd listens on the NNTP port (119), so either innd(8) has to be started on another port or the -p parameter used. Note that with this parameter, nnrpd continues running until killed. This means that it reads inn.conf once on startup and never again until restarted. nnrpd should therefore be restarted if inn.conf is changed.

When started in daemon mode, nnrpd will write its PID into a file in the pathrun directory. The file will be named nnrpd.pid if nnrpd listens on port 119 (default), or nnrpd-%d.pid, where %d is replaced with the port that nnrpd is configured to listen on (-p option is given and its argument is not 119).

-f

If specified, nnrpd does not detach itself and runs in the foreground when started as a standalone daemon using the -D flag.

-i initial

Specify an initial command to nnrpd. When used, initial is taken as if it were the first command received by nnrpd. After having responded, nnrpd will close the connection.

-I instance

If specified, instance is used as an additional static portion within message-IDs generated by nnrpd; typically this option would be used where a cluster of machines exist with the same virtual hostname and must be disambiguated during posts.

-n

The -n flag turns off resolution of IP addresses to names. If you only use IP-based restrictions in readers.conf and can handle IP addresses in your logs, using this flag may result in some additional speed.

-o

The -o flag causes all articles to be spooled instead of sending them to innd(8). rnews with the -U flag should be invoked from cron on a regular basis to take care of these articles. This flag is useful if innd(8) is accepting articles and nnrpd is started standalone or using inetd(8).

-p port

The -p parameter instructs nnrpd to listen on port when started as a standalone daemon using the -D flag.

-P prefork

The -P parameter instructs nnrpd to prefork prefork children awaiting connections when started as a standalone daemon using the -D flag.

-r reason

If the -r flag is used, then nnrpd will reject the incoming connection giving reason as the text. This flag is used by innd(8) when it is paused or throttled. reason should be encoded in UTF-8.

-s padding

As each command is received, nnrpd tries to change its "argv" array so that ps(1) will print out the command being executed. To get a full display, the -s flag may be used with a long string as its argument, which will be overwritten when the program changes its title.

-S

If specified, nnrpd will start a negotiation for a TLS session as soon as connected. To use this flag, the OpenSSL SSL and crypto libraries must have been found at configure time, or --with-openssl specified at configure time. For more information on running nnrpd with TLS support, see "TLS SUPPORT".

-t

If the -t flag is used, then all client commands and initial responses will be traced by reporting them in syslog. This flag is set by innd(8) under the control of the ctlinnd(8) "trace" command, and is toggled upon receipt of a SIGHUP; see signal(2).

TLS SUPPORT

If you're happy with a self-signed certificate (which will generate warnings with some news reader clients), you can create and install one in the default path by running "make cert" after "make install" when installing INN, or by running the following commands:

    umask 077
    openssl req -new -x509 -nodes -out <pathetc>/cert.pem \
        -days 366 -keyout <pathetc>/key.pem
    chown news:news <pathetc>/cert.pem
    chmod 640 <pathetc>/cert.pem
    chown news:news <pathetc>/key.pem
    chmod 600 <pathetc>/key.pem

Replace the paths with something appropriate to your INN installation. This will create a self-signed certificate that will expire in a year. The openssl program will ask you a variety of questions about your organization. Enter the fully qualified domain name of your news service (either the server canonical name or a dedicated alias for the news service) as the name the certificate is for.

You then have to set these inn.conf parameters with the right paths:

    tlscapath:      <pathetc>
    tlscertfile:    <pathetc>/cert.pem
    tlskeyfile:     <pathetc>/key.pem

If you want to use a complete certificate chain, you can directly put it in tlscertfile (like Apache's SSLCertificateFile directive). Alternately, you can put a single certificate in tlscertfile and use tlscafile for additional certificates needed to complete the chain, like a separate authority root certificate.

More concretly, when using Let's Encrypt certificates, Certbot's files can be installed as follows:

    tlscapath:      /etc/letsencrypt/live/news.server.com
    tlscertfile:    /etc/letsencrypt/live/news.server.com/fullchain.pem
    tlskeyfile:     /etc/letsencrypt/live/news.server.com/privkey.pem

or:

    tlscapath:      /etc/letsencrypt/live/news.server.com
    tlscafile:      /etc/letsencrypt/live/news.server.com/chain.pem
    tlscertfile:    /etc/letsencrypt/live/news.server.com/cert.pem
    tlskeyfile:     /etc/letsencrypt/live/news.server.com/privkey.pem

Make sure that the permission rights are properly set so that the news user or the news group can read these directories and files (typically, he should access /etc/letsencrypt/live/news.server.com and /etc/letsencrypt/archive/news.server.com where the real keys are located, and the private key should not be world-readable).

There are two common ways for a news client to negotiate a TLS connection: either via the use of a dedicated port (usually 563) on which TLS is immediately negotiated upon connection, or via the now discouraged way (per RFC 8143) to use the STARTTLS command on the usual NNTP port (119) to dynamically upgrade from unencrypted to TLS-protected traffic during an NNTP session. innd does not, however, know how to listen for connections to that separate port (563). You will therefore need to arrange for nnrpd to listen on that port through some other means. This can be done with the -D flag along with "-p 563" and put into your init scripts:

    su news -s /bin/sh -c '<pathbin>/nnrpd -D -p 563 -S'

but the easiest way is probably to add a line like:

    nntps stream tcp nowait news <pathbin>/nnrpd nnrpd -S

to /etc/inetd.conf or the equivalent on your system and let inetd run nnrpd. (Change the path to nnrpd to match your installation.) You may need to replace "nntps" with 563 if "nntps" isn't defined in /etc/services on your system.

Optionally, you may set the tlsciphers, tlsciphers13, tlscompression, tlseccurve, tlspreferserverciphers, and tlsprotocols parameters in inn.conf to fine-tune the behaviour of the TLS/SSL negotiation whenever a new attack on the TLS protocol or some supported cipher suite is discovered.

PROTOCOL DIFFERENCES

1.

The XGTITLE [wildmat] command is provided. This extension is used by ANU-News and documented in RFC 2980. It returns a 282 reply code, followed by a one-line description of all newsgroups that match the pattern. The default is the current group.

Note that LIST NEWSGROUPS should be used instead of XGTITLE.

2.

The XHDR header [message-ID|range] command is implemented. It returns a 221 reply code, followed by specific header fields for the specified range; the default is to return the data for the current article. See RFC 2980.

Note that HDR should be used instead of XHDR.

3.

The XOVER [range] command is provided. It returns a 224 reply code, followed by the overview data for the specified range; the default is to return the data for the current article. See RFC 2980.

Note that OVER should be used instead of XOVER.

4.

A new command, XPAT header message-ID|range pattern [pattern ...], is provided. The first argument is the case-insensitive name of the header field to be searched. The second argument is either an article range or a single message-ID, as specified in RFC 2980. The third argument is a uwildmat-style pattern; if there are additional arguments, they are joined together separated by a single space to form the complete pattern. This command is similar to the XHDR command. It returns a 221 response code, followed by the text response of all article numbers that match the pattern.

5.

A newsgroup name is case-sensitive for nnrpd.

6.

If IHAVE has been advertised, it will not necessarily be advertised for the entire session (contrary to section 3.4.1 of RFC 3977). nnrpd only advertises the IHAVE capability when it is really available.

7.

nnrpd allows a wider syntax for wildmats and ranges (especially "-" and "-article-number").

HISTORY

SEE ALSO