gnunet/doc/handbook/chapters/installation.texi

@node Installing GNUnet
@chapter Installing GNUnet

This guide is intended for those who want to install Gnunet from
source. For instructions on how to install GNUnet as a binary package
please refer to the official documentation of your operating system or
package manager.

For understanding this guide properly it is important to know that
there are two different ways of running GNUnet:

@itemize @bullet
@item the @emph{single-user setup}
@item the @emph{multi-user setup}
@end itemize

The latter variant has a better security model and requires extra
preparation before running @code{make install} and a different
configuration. Beginners who want to quickly try out GNUnet can
use the @emph{single-user setup}.

@menu
* Installing dependencies::
* Getting the Source Code::
* Create user and groups for the system services::
* Preparing and Compiling the Source Code::
* Installation::
* Minimal configuration::
* Checking the Installation::
* The graphical configuration interface::
* Config Leftovers::
@end menu

@c -----------------------------------------------------------------------
@node Installing dependencies
@section Installing dependencies

GNUnet needs few libraries and applications for being able to run and
another few optional ones for using certain features. Preferably they
should be installed with a package manager.

The mandatory libraries and applications are
@itemize @bullet
@item autoconf 2.59 or above
@item automake 1.11.1 or above
@item gettext
@item glibc (read below, other libcs work)
@item GnuTLS 3.2.12 or above, recommended to be linked against libunbound
@item GNU make 4.0 or higher (other make implementations do work)
@item iptables (on Linux systems)
@item libtool 2.2 or above
@item libltdl (part of libtool)
@item libgcrypt 1.6 or above
@item libidn2 or libidn
@item libmicrohttpd 0.9.63 or above
@item libunistring
@item libgmp
@item libgnurl or libcurl (libcurl has to be linked to GnuTLS) 7.35.0 or above
@item Texinfo 5.2 or above (for building the documentation)
@item Texlive 2012 or above (for building the documentation, and for gnunet-bcd)
@item makeinfo 4.8 or above
@item pkgconf (or pkg-config)
@item zlib
@end itemize

Glibc is required for certain NSS features:

@example
One mechanism of integrating GNS with legacy applications via NSS is
not available if this is disabled. But applications that don't use the
glibc for NS resolution won't work anyway with this, so little is lost
on *BSD systems.
GNS via direct use or via the HTTP or DNS proxies is unaffected.
@end example

Other libcs should work, the resulting builds just don't include the
glibc NSS specific code. One example is the build against NetBSD's libc
as detailed in @uref{https://bugs.gnunet.org/view.php?id=5605}.

In addition GNUnet needs at least one of these three databases
(at the minimum sqlite3)
@itemize @bullet
@item sqlite + libsqlite 3.8 or above (the default, requires no further configuration)
@item postgres + libpq
@item mysql + libmysqlclient
@end itemize

These are the dependencies only required for certain features
@itemize @bullet
@item miniupnpc (for traversing NAT boxes more reliably)
@item libnss
@item libopus (for running the GNUnet conversation telephony application)
@item libogg (for running the GNUnet conversation telephony application)
@item gstreamer OR libpulse (for running the GNUnet conversation telephony application)
@item bluez (for bluetooth support)
@item libextractor (optional but highly recommended, read below)
@item libpbc
(for attribute-based encryption and the identity provider subsystem)
@item libgabe
(for attribute-based encryption and the identity provider subsystem)
@item texi2mdoc (for automatic mdoc generation)
@item perl5 for some utilities (which are not installed)
@item libjanson
@end itemize

About libextractor being optional:
@example
While libextractor ("LE") is optional, it is recommended to build gnunet
against it. If you install it later, you won't benefit from libextractor.
If you are a distributor, we recommend to split LE into basis + plugins
rather than making LE an option as an afterthought by the user.  LE
itself is very small, but its dependency chain on first, second, third
etc level can be big.  There is a small effect on privacy if your LE
build differs from one which includes all plugins (plugins are build as
shared objects): if users publish a directory with a mixture of file
types (for example mpeg, jpeg, png, gif) the configuration of LE could
leak which plugins are installed for which filetypes are not providing
more details.  However, this leak is just a minor concern.
@end example

These are the test-suite requirements:
@itemize @bullet
@item python3.6 or higher
@item gnunet (installation first)
@item some core-utils: which(1), bc(1), curl(1), sed(1), awk(1), etc.
@item a shell (very few Bash scripts, the majority are POSIX sh scripts)
@end itemize

These are runtime requirements:
@itemize @bullet
@item nss (the certutil binary, for gnunet-gns-proxy-setup-ca)
@item openssl (openssl binary, for gnunet-gns-proxy-setup-ca)
@end itemize

@c -----------------------------------------------------------------------
@node Getting the Source Code
@section Getting the Source Code
You can either download the source code using git (you obviously need
git installed) or as an archive.

Using git type
@example
git clone https://git.gnunet.org/gnunet.git
@end example

The archive can be found at
@uref{https://ftpmirror.gnu.org/gnu/gnunet/}. Extract it using a graphical
archive tool or @code{tar}:
@example
tar xzvf gnunet-@value{VERSION}.tar.gz
@end example

In the next chapter we will assume that the source code is available
in the home directory at @code{~/gnunet}.

@c -----------------------------------------------------------------------
@node Create user and groups for the system services
@section Create user and groups for the system services

@cartouche
For the single-user setup this section can be skipped.
@end cartouche

The multi-user setup means that there are @emph{system services}, which are
run once per machine as a dedicated system user (called @code{gnunet}) and
@emph{user services} which can be started by every user who wants to use
GNUnet applications. The user services communicate with the system services
over unix domain sockets. To gain permissions to read and write those sockets
the users running GNUnet applications will need to be in the @code{gnunet}
group. In addition the group @code{gnunetdns} may be needed (see below).

Create user @code{gnunet} who is member of the group @code{gnunet}
(automatically created) and specify a home directory where the GNUnet
services will store persistent data such as information about peers.
@example
$ sudo useradd --system --home-dir /var/lib/gnunet --create-home gnunet
@end example

Now add your own user to the @code{gnunet} group.

@example
$ sudo usermod -aG gnunet alice
@end example

Create a group @code{gnunetdns}. This allows using @code{setgid} in a way
that only the DNS service can run the @code{gnunet-helper-dns} binary. This
is only needed if @emph{system-wide DNS interception} will be used. For more
information see @xref{Configuring system-wide DNS interception}.

@example
$ sudo groupadd gnunetdns
@end example

@c -----------------------------------------------------------------------
@node Preparing and Compiling the Source Code
@section Preparing and Compiling the Source Code
For preparing the source code for compilation a bootstrap script and
@code{configure} has to be run from the source code directory. When
running @code{configure} the following options can be specified to
customize the compilation and installation process:

@itemize @bullet
@item @code{--disable-documentation} - don't build the documentation
@item @code{--enable-logging=[LOGLEVEL]} - choose a loglevel (@code{debug}, @code{info}, @code{warning} or @code{error})
@item @code{--prefix=[PATH]} - the directory where the GNUnet libraries and binaries will be installed
@item @code{--with-extractor=[PATH]} - the path to libextractor
@item @code{--with-libidn=[PATH]} - the path to libidn
@item @code{--with-libidn2=[PATH]} - the path to libidn2 (takes priority over libidn if both are found)
@item @code{--with-microhttpd=[PATH]} - the path to libmicrohttpd
@item @code{--with-sqlite=[PATH]} - the path to libsqlite
@item @code{--with-zlib=[PATH]} - the path to zlib
@end itemize

Note that the list above is not always up to date and you
should check the output of @code{./configure --help}, read
the @file{configure.ac} or send an email asking for assistance
if you are in doubt of any configure options or require fixes
for your operating system.

The following example configures the installation prefix
@code{/usr/local} and disables building the documentation
@example
$ cd ~/gnunet
$ ./bootstrap
$ configure --prefix=/usr/local --disable-documentation
@end example

After running the bootstrap script and @code{configure} successfully
the source code can be compiled with make. Here @code{-j5} specifies
that 5 threads should be used.
@example
$ make -j5
@end example

@c -----------------------------------------------------------------------
@node Installation
@section Installation
The compiled binaries can be installed using @code{make install}. It
needs to be run as root (or with sudo) because some binaries need the
@code{suid} bit set. Without that some features (e.g. the VPN service,
system-wide DNS interception, NAT traversal using ICMP) will not work.

@example
$ sudo make install
@end example

@menu
* NSS plugin (Optional)::
* Installing the GNS Certificate Authority (Optional)::
@end menu

@node NSS plugin (Optional)
@subsection NSS plugin (Optional)

@cartouche
The installation of the NSS plugin is only necessary if GNS
resolution shall be used with legacy applications (that only
support DNS).
@end cartouche

One important library is the GNS plugin for NSS (the name services
switch) which allows using GNS (the GNU name system) in the normal DNS
resolution process. Unfortunately NSS expects it in a specific
location (probably @code{/lib}) which may differ from the installation
prefix (see @code{--prefix} option in the previous section). This is
why the plugin has to be installed manually.

Find the directory where nss plugins are installed on your system, e.g.

@example
$ ls -l /lib/libnss_*
/lib/libnss_mymachines.so.2
/lib/libnss_resolve.so.2
/lib/libnss_myhostname.so.2
/lib/libnss_systemd.so.2
@end example

Copy the GNS NSS plugin to that directory:

@example
cp ~/gnunet/src/gns/nss/.libs/libnss_gns.so.2 /lib
@end example

Now, to activate the plugin, you need to edit your
@code{/etc/nsswitch.conf} where you should find a line like this:

@example
hosts: files mdns4_minimal [NOTFOUND=return] dns mdns4
@end example

The exact details may differ a bit, which is fine. Add the text
@code{"gns [NOTFOUND=return]"} after @code{"files"}.

@example
hosts: files gns [NOTFOUND=return] mdns4_minimal [NOTFOUND=return] dns mdns4
@end example

@node Installing the GNS Certificate Authority (Optional)
@subsection Installing the GNS Certificate Authority (Optional)

@cartouche
Installing the GNS certificate authority is only necessary if GNS shall
be used in a browser.
@end cartouche

The GNS Certificate authority can provide TLS certificates for GNS names while
downloading webpages from legacy webservers. This allows browsers to use HTTPS
in combinations with GNS name resolution.

To install it execute the GNS CA-setup script. So far Firefox and Chromium are
supported.

@example
$ gnunet-gns-proxy-setup-ca
@end example

A local proxy server, that takes care of the name resolution and provides
certificates on-the-fly needs to be started:

@example
$ /usr/lib/gnunet/libexec/gnunet-gns-proxy
@end example

Now GNS should work in browsers that are configured to use a SOCKS proxy on
@code{localhost:7777}.


@node Minimal configuration
@section Minimal configuration
GNUnet needs a configuration file to start (@pxref{Config file format}).
For the @emph{single-user setup} an empty file is sufficient:

@example
$ touch ~/.config/gnunet.conf
@end example

For the @emph{multi-user setup} we need an extra config file for the system
services. The default location is @code{/etc/gnunet.conf}. The minimal
content of that file which activates the system services roll is:

@example
[arm]
START_SYSTEM_SERVICES = YES
START_USER_SERVICES = NO
@end example

The config file for the user services (@code{~/.config/gnunet.conf}) needs
the opposite configuration to activate the user services roll:

@example
[arm]
START_SYSTEM_SERVICES = NO
START_USER_SERVICES = YES
@end example


@node Checking the Installation
@section Checking the Installation


This section describes a quick, casual way to check if your GNUnet
installation works. However, if it does not, we do not cover
steps for recovery --- for this, please study the instructions
provided in the developer handbook as well as the system-specific
instruction in the source code repository.
Please note that the system specific instructions are not provided
as part of this handbook!


@menu
* Starting GNUnet::
* gnunet-gtk::
* Statistics::
* Peer Information::
@end menu

@cindex Starting GNUnet
@cindex GNUnet GTK
@cindex GTK
@cindex GTK user interface

@node Starting GNUnet
@subsection Starting GNUnet
The GNUnet services are started and stopped by the ARM service (Automatic
Restart Manager). For the @emph{single-user setup} a simple

@example
$ gnunet-arm -s
@end example

starts a default set of services. Later GNUnet applications can request more
services to start without additional user interaction. GNUnet can be stopped
again using the @code{-e} option:

@example
$ gnunet-arm -e
@end example

The list of running services can be displayed using the @code{-I} option.
It should look similar to this example:

@example
$ gnunet-arm -I
Running services:
topology (gnunet-daemon-topology)
nat (gnunet-service-nat)
vpn (gnunet-service-vpn)
gns (gnunet-service-gns)
cadet (gnunet-service-cadet)
namecache (gnunet-service-namecache)
hostlist (gnunet-daemon-hostlist)
revocation (gnunet-service-revocation)
ats (gnunet-service-ats)
peerinfo (gnunet-service-peerinfo)
zonemaster (gnunet-service-zonemaster)
zonemaster-monitor (gnunet-service-zonemaster-monitor)
dht (gnunet-service-dht)
namestore (gnunet-service-namestore)
set (gnunet-service-set)
statistics (gnunet-service-statistics)
nse (gnunet-service-nse)
fs (gnunet-service-fs)
peerstore (gnunet-service-peerstore)
core (gnunet-service-core)
rest (gnunet-rest-server)
transport (gnunet-service-transport)
datastore (gnunet-service-datastore)
@end example

For the @emph{multi-user setup} first the system services need to be started
as the system user, i.e. the user @code{gnunet} needs to execute
@code{gnunet-arm -s}. This should be done by the system's init system.
Then the user who wants to start GNUnet applications has to run
@code{gnunet-arm -s} too. It is recommended to automate this, e.g. using
the user's crontab.

@node gnunet-gtk
@subsection gnunet-gtk


The @command{gnunet-gtk} package contains several graphical
user interfaces for the respective GNUnet applications.
Currently these interfaces cover:

@itemize @bullet
@item Statistics
@item Peer Information
@item GNU Name System
@item File Sharing
@item Conversation
@item Setup
@end itemize

Previously, many of these interfaces were combined into one application
called @command{gnunet-gtk}, with different tabs for each interface. This
combined application has been removed in version 0.11.0, but each of the
interfaces is still available as a standalone application
(@command{gnunet-statistics-gtk} for statistics, @command{gnunet-fs-gtk}
for filesharing, etc).

@node Statistics
@subsection Statistics


We assume that you have started gnunet via @code{gnunet-arm} or via your
system-provided method for starting services.
First, you should launch GNUnet's graphical statistics interface.
You can do this from the command-line by typing

@example
gnunet-statistics-gtk
@end example

If your peer is running correctly, you should see a bunch
of lines, all of which should be ``significantly'' above zero (at
least if your peer has been running for more than a few seconds). The
lines indicate how many other peers your peer is connected to (via
different mechanisms) and how large the entire overlay network is
currently estimated to be. The X-axis represents time (in seconds
since the start of @command{gnunet-statistics-gtk}).

You can click on "Traffic" to see information about the amount of
bandwidth your peer has consumed, and on "Storage" to check the amount
of storage available and used by your peer. Note that "Traffic" is
plotted cumulatively, so you should see a strict upwards trend in the
traffic.

The term ``peer'' is a common word used in
federated and distributed networks to describe a participating device
which is connected to the network. Thus, your Personal Computer or
whatever it is you are looking at the Gtk+ interface describes a
``Peer'' or a ``Node''.

@node Peer Information
@subsection Peer Information


First, you should launch the peer information graphical user interface.
You can do this from the command-line by typing

@example
$ gnunet-peerinfo-gtk
@end example

Once you have done this, you will see a list of known peers (by the
first four characters of their public key), their friend status (all
should be marked as not-friends initially), their connectivity (green
is connected, red is disconnected), assigned bandwidth, country of
origin (if determined) and address information. If hardly any peers
are listed and/or if there are very few peers with a green light for
connectivity, there is likely a problem with your network
configuration.

@c NOTE: Inserted from Installation Handbook in original ``order'':
@c FIXME: Move this to User Handbook.
@node The graphical configuration interface
@section The graphical configuration interface

If you also would like to use @command{gnunet-gtk} and
@command{gnunet-setup} (highly recommended for beginners), do:

@menu
* Configuring your peer::
* Configuring the Friend-to-Friend (F2F) mode::
* Configuring the hostlist to bootstrap::
* Configuration of the HOSTLIST proxy settings::
* Configuring your peer to provide a hostlist ::
* Configuring the datastore::
* Configuring the MySQL database::
* Reasons for using MySQL::
* Reasons for not using MySQL::
* Setup Instructions::
* Testing::
* Performance Tuning::
* Setup for running Testcases::
* Configuring the Postgres database::
* Reasons to use Postgres::
* Reasons not to use Postgres::
* Manual setup instructions::
* Testing the setup manually::
* Configuring the datacache::
* Configuring the file-sharing service::
* Configuring logging::
* Configuring the transport service and plugins::
* Configuring the WLAN transport plugin::
* Configuring HTTP(S) reverse proxy functionality using Apache or nginx::
* Blacklisting peers::
* Configuration of the HTTP and HTTPS transport plugins::
* Configuring the GNU Name System::
* Configuring the GNUnet VPN::
* Bandwidth Configuration::
* Configuring NAT::
* Peer configuration for distributors (e.g. Operating Systems)::
@end menu

@node Configuring your peer
@subsection Configuring your peer

This chapter will describe the various configuration options in GNUnet.

The easiest way to configure your peer is to use the
@command{gnunet-setup} tool.
@command{gnunet-setup} is part of the @command{gnunet-gtk}
package. You might have to install it separately.

Many of the specific sections from this chapter actually are linked from
within @command{gnunet-setup} to help you while using the setup tool.

While you can also configure your peer by editing the configuration
file by hand, this is not recommended for anyone except for developers
as it requires a more in-depth understanding of the configuration files
and internal dependencies of GNUnet.

@node Configuring the Friend-to-Friend (F2F) mode
@subsection Configuring the Friend-to-Friend (F2F) mode

GNUnet knows three basic modes of operation:
@itemize @bullet
@item In standard "peer-to-peer" mode,
your peer will connect to any peer.
@item In the pure "friend-to-friend"
mode, your peer will ONLY connect to peers from a list of friends
specified in the configuration.
@item Finally, in mixed mode,
GNUnet will only connect to arbitrary peers if it
has at least a specified number of connections to friends.
@end itemize

When configuring any of the F2F ("friend-to-friend") modes,
you first need to create a file with the peer identities
of your friends. Ask your friends to run

@example
$ gnunet-peerinfo -sq
@end example

@noindent
The resulting output of this command needs to be added to your
@file{friends} file, which is simply a plain text file with one line
per friend with the output from the above command.

You then specify the location of your @file{friends} file in the
@code{FRIENDS} option of the "topology" section.

Once you have created the @file{friends} file, you can tell GNUnet to only
connect to your friends by setting the @code{FRIENDS-ONLY} option
(again in the "topology" section) to YES.

If you want to run in mixed-mode, set "FRIENDS-ONLY" to NO and configure a
minimum number of friends to have (before connecting to arbitrary peers)
under the "MINIMUM-FRIENDS" option.

If you want to operate in normal P2P-only mode, simply set
@code{MINIMUM-FRIENDS} to zero and @code{FRIENDS_ONLY} to NO.
This is the default.

@node Configuring the hostlist to bootstrap
@subsection Configuring the hostlist to bootstrap

After installing the software you need to get connected to the GNUnet
network. The configuration file included in your download is already
configured to connect you to the GNUnet network.
In this section the relevant configuration settings are explained.

To get an initial connection to the GNUnet network and to get to know
peers already connected to the network you can use the so called
"bootstrap servers".
These servers can give you a list of peers connected to the network.
To use these bootstrap servers you have to configure the hostlist daemon
to activate bootstrapping.

To activate bootstrapping, edit the @code{[hostlist]}-section in your
configuration file. You have to set the argument @command{-b} in the
options line:

@example
[hostlist]
OPTIONS = -b
@end example

Additionally you have to specify which server you want to use.
The default bootstrapping server is
"@uref{http://v10.gnunet.org/hostlist, http://v10.gnunet.org/hostlist}".
[^] To set the server you have to edit the line "SERVERS" in the hostlist
section. To use the default server you should set the lines to

@example
SERVERS = http://v10.gnunet.org/hostlist [^]
@end example

@noindent
To use bootstrapping your configuration file should include these lines:

@example
[hostlist]
OPTIONS = -b
SERVERS = http://v10.gnunet.org/hostlist [^]
@end example

@noindent
Besides using bootstrap servers you can configure your GNUnet peer to
receive hostlist advertisements.
Peers offering hostlists to other peers can send advertisement messages
to peers that connect to them. If you configure your peer to receive these
messages, your peer can download these lists and connect to the peers
included. These lists are persistent, which means that they are saved to
your hard disk regularly and are loaded during startup.

To activate hostlist learning you have to add the @command{-e}
switch to the @code{OPTIONS} line in the hostlist section:

@example
[hostlist]
OPTIONS = -b -e
@end example

@noindent
Furthermore you can specify in which file the lists are saved.
To save the lists in the file @file{hostlists.file} just add the line:

@example
HOSTLISTFILE = hostlists.file
@end example

@noindent
Best practice is to activate both bootstrapping and hostlist learning.
So your configuration file should include these lines:

@example
[hostlist]
OPTIONS = -b -e
HTTPPORT = 8080
SERVERS = http://v10.gnunet.org/hostlist [^]
HOSTLISTFILE = $SERVICEHOME/hostlists.file
@end example

@node Configuration of the HOSTLIST proxy settings
@subsection Configuration of the HOSTLIST proxy settings

The hostlist client can be configured to use a proxy to connect to the
hostlist server.
This functionality can be configured in the configuration file directly
or using the @command{gnunet-setup} tool.

The hostlist client supports the following proxy types at the moment:

@itemize @bullet
@item HTTP and HTTP 1.0 only proxy
@item SOCKS 4/4a/5/5 with hostname
@end itemize

In addition authentication at the proxy with username and password can be
configured.

To configure proxy support for the hostlist client in the
@command{gnunet-setup} tool, select the "hostlist" tab and select
the appropriate proxy type.
The hostname or IP address (including port if required) has to be entered
in the "Proxy hostname" textbox. If required, enter username and password
in the "Proxy username" and "Proxy password" boxes.
Be aware that this information will be stored in the configuration in
plain text (TODO: Add explanation and generalize the part in Chapter 3.6
about the encrypted home).

To provide these options directly in the configuration, you can
enter the following settings in the @code{[hostlist]} section of
the configuration:

@example
# Type of proxy server,
# Valid values: HTTP, HTTP_1_0, SOCKS4, SOCKS5, SOCKS4A, SOCKS5_HOSTNAME
# Default: HTTP
# PROXY_TYPE = HTTP

# Hostname or IP of proxy server
# PROXY =
# User name for proxy server
# PROXY_USERNAME =
# User password for proxy server
# PROXY_PASSWORD =
@end example

@node Configuring your peer to provide a hostlist
@subsection Configuring your peer to provide a hostlist

If you operate a peer permanently connected to GNUnet you can configure
your peer to act as a hostlist server, providing other peers the list of
peers known to him.

Your server can act as a bootstrap server and peers needing to obtain a
list of peers can contact it to download this list.
To download this hostlist the peer uses HTTP.
For this reason you have to build your peer with libgnurl (or libcurl)
and microhttpd support.

To configure your peer to act as a bootstrap server you have to add the
@command{-p} option to @code{OPTIONS} in the @code{[hostlist]} section
of your configuration file.
Besides that you have to specify a port number for the http server.
In conclusion you have to add the following lines:

@example
[hostlist]
HTTPPORT = 12980
OPTIONS = -p
@end example

@noindent
If your peer acts as a bootstrap server other peers should know about
that. You can advertise the hostlist your are providing to other peers.
Peers connecting to your peer will get a message containing an
advertisement for your hostlist and the URL where it can be downloaded.
If this peer is in learning mode, it will test the hostlist and, in the
case it can obtain the list successfully, it will save it for
bootstrapping.

To activate hostlist advertisement on your peer, you have to set the
following lines in your configuration file:

@example
[hostlist]
EXTERNAL_DNS_NAME = example.org
HTTPPORT = 12981
OPTIONS = -p -a
@end example

@noindent
With this configuration your peer will a act as a bootstrap server and
advertise this hostlist to other peers connecting to it.
The URL used to download the list will be
@code{@uref{http://example.org:12981/, http://example.org:12981/}}.

Please notice:

@itemize @bullet
@item The hostlist is @b{not} human readable, so you should not try to
download it using your webbrowser. Just point your GNUnet peer to the
address!
@item Advertising without providing a hostlist does not make sense and
will not work.
@end itemize

@node Configuring the datastore
@subsection Configuring the datastore

The datastore is what GNUnet uses for long-term storage of file-sharing
data. Note that long-term does not mean 'forever' since content does have
an expiration date, and of course storage space is finite (and hence
sometimes content may have to be discarded).

Use the @code{QUOTA} option to specify how many bytes of storage space
you are willing to dedicate to GNUnet.

In addition to specifying the maximum space GNUnet is allowed to use for
the datastore, you need to specify which database GNUnet should use to do
so. Currently, you have the choice between sqLite, MySQL and Postgres.

@node Configuring the MySQL database
@subsection Configuring the MySQL database

This section describes how to setup the MySQL database for GNUnet.

Note that the mysql plugin does NOT work with mysql before 4.1 since we
need prepared statements.
We are generally testing the code against MySQL 5.1 at this point.

@node Reasons for using MySQL
@subsection Reasons for using MySQL

@itemize @bullet

@item On up-to-date hardware where
mysql can be used comfortably, this module
will have better performance than the other database choices (according
to our tests).

@item Its often possible to recover the mysql database from internal
inconsistencies. Some of the other databases do not support repair.
@end itemize

@node Reasons for not using MySQL
@subsection Reasons for not using MySQL

@itemize @bullet
@item Memory usage (likely not an issue if you have more than 1 GB)
@item Complex manual setup
@end itemize

@node Setup Instructions
@subsection Setup Instructions

@itemize @bullet

@item In @file{gnunet.conf} set in section @code{DATASTORE} the value for
@code{DATABASE} to @code{mysql}.

@item Access mysql as root:

@example
$ mysql -u root -p
@end example

@noindent
and issue the following commands, replacing $USER with the username
that will be running @command{gnunet-arm} (so typically "gnunet"):

@example
CREATE DATABASE gnunet;
GRANT select,insert,update,delete,create,alter,drop,create \
temporary tables ON gnunet.* TO $USER@@localhost;
SET PASSWORD FOR $USER@@localhost=PASSWORD('$the_password_you_like');
FLUSH PRIVILEGES;
@end example

@item
In the $HOME directory of $USER, create a @file{.my.cnf} file with the
following lines

@example
[client]
user=$USER
password=$the_password_you_like
@end example

@end itemize

That's it. Note that @file{.my.cnf} file is a slight security risk unless
its on a safe partition. The @file{$HOME/.my.cnf} can of course be
a symbolic link.
Luckily $USER has only privileges to mess up GNUnet's tables,
which should be pretty harmless.

@node Testing
@subsection Testing

You should briefly try if the database connection works. First, login
as $USER. Then use:

@example
$ mysql -u $USER
mysql> use gnunet;
@end example

@noindent
If you get the message

@example
Database changed
@end example

@noindent
it probably works.

If you get

@example
ERROR 2002: Can't connect to local MySQL server
through socket '/tmp/mysql.sock' (2)
@end example

@noindent
it may be resolvable by

@example
ln -s /var/run/mysqld/mysqld.sock /tmp/mysql.sock
@end example

@noindent
so there may be some additional trouble depending on your mysql setup.

@node Performance Tuning
@subsection Performance Tuning

For GNUnet, you probably want to set the option

@example
innodb_flush_log_at_trx_commit = 0
@end example

@noindent
for a rather dramatic boost in MySQL performance. However, this reduces
the "safety" of your database as with this options you may loose
transactions during a power outage.
While this is totally harmless for GNUnet, the option applies to all
applications using MySQL. So you should set it if (and only if) GNUnet is
the only application on your system using MySQL.

@node Setup for running Testcases
@subsection Setup for running Testcases

If you want to run the testcases, you must create a second database
"gnunetcheck" with the same username and password. This database will
then be used for testing (@command{make check}).

@node Configuring the Postgres database
@subsection Configuring the Postgres database

This text describes how to setup the Postgres database for GNUnet.

This Postgres plugin was developed for Postgres 8.3 but might work for
earlier versions as well.

@node Reasons to use Postgres
@subsection Reasons to use Postgres

@itemize @bullet
@item Easier to setup than MySQL
@item Real database
@end itemize

@node Reasons not to use Postgres
@subsection Reasons not to use Postgres

@itemize @bullet
@item Quite slow
@item Still some manual setup required
@end itemize

@node Manual setup instructions
@subsection Manual setup instructions

@itemize @bullet
@item In @file{gnunet.conf} set in section @code{DATASTORE} the value for
@code{DATABASE} to @code{postgres}.
@item Access Postgres to create a user:

@table @asis
@item with Postgres 8.x, use:

@example
# su - postgres
$ createuser
@end example

@noindent
and enter the name of the user running GNUnet for the role interactively.
Then, when prompted, do not set it to superuser, allow the creation of
databases, and do not allow the creation of new roles.

@item with Postgres 9.x, use:

@example
# su - postgres
$ createuser -d $GNUNET_USER
@end example

@noindent
where $GNUNET_USER is the name of the user running GNUnet.

@end table


@item
As that user (so typically as user "gnunet"), create a database (or two):

@example
$ createdb gnunet
# this way you can run "make check"
$ createdb gnunetcheck
@end example

@end itemize

Now you should be able to start @code{gnunet-arm}.

@node Testing the setup manually
@subsection Testing the setup manually

You may want to try if the database connection works. First, again login
as the user who will run @command{gnunet-arm}. Then use:

@example
$ psql gnunet # or gnunetcheck
gnunet=> \dt
@end example

@noindent
If, after you have started @command{gnunet-arm} at least once, you get
a @code{gn090} table here, it probably works.

@node Configuring the datacache
@subsection Configuring the datacache


The datacache is what GNUnet uses for storing temporary data. This data is
expected to be wiped completely each time GNUnet is restarted (or the
system is rebooted).

You need to specify how many bytes GNUnet is allowed to use for the
datacache using the @code{QUOTA} option in the section @code{[dhtcache]}.
Furthermore, you need to specify which database backend should be used to
store the data. Currently, you have the choice between
sqLite, MySQL and Postgres.

@node Configuring the file-sharing service
@subsection Configuring the file-sharing service

In order to use GNUnet for file-sharing, you first need to make sure
that the file-sharing service is loaded.
This is done by setting the @code{START_ON_DEMAND} option in
section @code{[fs]} to "YES". Alternatively, you can run

@example
$ gnunet-arm -i fs
@end example

@noindent
to start the file-sharing service by hand.

Except for configuring the database and the datacache the only important
option for file-sharing is content migration.

Content migration allows your peer to cache content from other peers as
well as send out content stored on your system without explicit requests.
This content replication has positive and negative impacts on both system
performance and privacy.

FIXME: discuss the trade-offs. Here is some older text about it...

Setting this option to YES allows gnunetd to migrate data to the local
machine. Setting this option to YES is highly recommended for efficiency.
Its also the default. If you set this value to YES, GNUnet will store
content on your machine that you cannot decrypt.
While this may protect you from liability if the judge is sane, it may
not (IANAL). If you put illegal content on your machine yourself, setting
this option to YES will probably increase your chances to get away with it
since you can plausibly deny that you inserted the content.
Note that in either case, your anonymity would have to be broken first
(which may be possible depending on the size of the GNUnet network and the
strength of the adversary).

@node Configuring logging
@subsection Configuring logging

Since version 0.9.0, logging in GNUnet is controlled via the
@code{-L} and @code{-l} options.
Using @code{-L}, a log level can be specified. With log level
@code{ERROR} only serious errors are logged.
The default log level is @code{WARNING} which causes anything of
concern to be logged.
Log level @code{INFO} can be used to log anything that might be
interesting information whereas
@code{DEBUG} can be used by developers to log debugging messages
(but you need to run @code{./configure} with
@code{--enable-logging=verbose} to get them compiled).
The @code{-l} option is used to specify the log file.

Since most GNUnet services are managed by @code{gnunet-arm}, using the
@code{-l} or @code{-L} options directly is not possible.
Instead, they can be specified using the @code{OPTIONS} configuration
value in the respective section for the respective service.
In order to enable logging globally without editing the @code{OPTIONS}
values for each service, @command{gnunet-arm} supports a
@code{GLOBAL_POSTFIX} option.
The value specified here is given as an extra option to all services for
which the configuration does contain a service-specific @code{OPTIONS}
field.

@code{GLOBAL_POSTFIX} can contain the special sequence "@{@}" which
is replaced by the name of the service that is being started.
Furthermore, @code{GLOBAL_POSTFIX} is special in that sequences
starting with "$" anywhere in the string are expanded (according
to options in @code{PATHS}); this expansion otherwise is
only happening for filenames and then the "$" must be the
first character in the option. Both of these restrictions do
not apply to @code{GLOBAL_POSTFIX}.
Note that specifying @code{%} anywhere in the @code{GLOBAL_POSTFIX}
disables both of these features.

In summary, in order to get all services to log at level
@code{INFO} to log-files called @code{SERVICENAME-logs}, the
following global prefix should be used:

@example
GLOBAL_POSTFIX = -l $SERVICEHOME/@{@}-logs -L INFO
@end example

@node Configuring the transport service and plugins
@subsection Configuring the transport service and plugins

The transport service in GNUnet is responsible to maintain basic
connectivity to other peers.
Besides initiating and keeping connections alive it is also responsible
for address validation.

The GNUnet transport supports more than one transport protocol.
These protocols are configured together with the transport service.

The configuration section for the transport service itself is quite
similar to all the other services

@example
START_ON_DEMAND = YES
@@UNIXONLY@@ PORT = 2091
HOSTNAME = localhost
HOME = $SERVICEHOME
CONFIG = $DEFAULTCONFIG
BINARY = gnunet-service-transport
#PREFIX = valgrind
NEIGHBOUR_LIMIT = 50
ACCEPT_FROM = 127.0.0.1;
ACCEPT_FROM6 = ::1;
PLUGINS = tcp udp
UNIXPATH = /tmp/gnunet-service-transport.sock
@end example

Different are the settings for the plugins to load @code{PLUGINS}.
The first setting specifies which transport plugins to load.

@itemize @bullet
@item transport-unix
A plugin for local only communication with UNIX domain sockets. Used for
testing and available on unix systems only. Just set the port

@example
[transport-unix]
PORT = 22086
TESTING_IGNORE_KEYS = ACCEPT_FROM;
@end example

@item transport-tcp
A plugin for communication with TCP. Set port to 0 for client mode with
outbound only connections

@example
[transport-tcp]
# Use 0 to ONLY advertise as a peer behind NAT (no port binding)
PORT = 2086
ADVERTISED_PORT = 2086
TESTING_IGNORE_KEYS = ACCEPT_FROM;
# Maximum number of open TCP connections allowed
MAX_CONNECTIONS = 128
@end example

@item transport-udp
A plugin for communication with UDP. Supports peer discovery using
broadcasts.

@example
[transport-udp]
PORT = 2086
BROADCAST = YES
BROADCAST_INTERVAL = 30 s
MAX_BPS = 1000000
TESTING_IGNORE_KEYS = ACCEPT_FROM;
@end example

@item transport-http
HTTP and HTTPS support is split in two part: a client plugin initiating
outbound connections and a server part accepting connections from the
client. The client plugin just takes the maximum number of connections as
an argument.

@example
[transport-http_client]
MAX_CONNECTIONS = 128
TESTING_IGNORE_KEYS = ACCEPT_FROM;
@end example

@example
[transport-https_client]
MAX_CONNECTIONS = 128
TESTING_IGNORE_KEYS = ACCEPT_FROM;
@end example

@noindent
The server has a port configured and the maximum number of connections.
The HTTPS part has two files with the certificate key and the certificate
file.

The server plugin supports reverse proxies, so a external hostname can be
set using the @code{EXTERNAL_HOSTNAME} setting.
The webserver under this address should forward the request to the peer
and the configure port.

@example
[transport-http_server]
EXTERNAL_HOSTNAME = fulcrum.net.in.tum.de/gnunet
PORT = 1080
MAX_CONNECTIONS = 128
TESTING_IGNORE_KEYS = ACCEPT_FROM;
@end example

@example
[transport-https_server]
PORT = 4433
CRYPTO_INIT = NORMAL
KEY_FILE = https.key
CERT_FILE = https.cert
MAX_CONNECTIONS = 128
TESTING_IGNORE_KEYS = ACCEPT_FROM;
@end example

@item transport-wlan

The next section describes how to setup the WLAN plugin,
so here only the settings. Just specify the interface to use:

@example
[transport-wlan]
# Name of the interface in monitor mode (typically monX)
INTERFACE = mon0
# Real hardware, no testing
TESTMODE = 0
TESTING_IGNORE_KEYS = ACCEPT_FROM;
@end example
@end itemize

@node Configuring the WLAN transport plugin
@subsection Configuring the WLAN transport plugin

The wlan transport plugin enables GNUnet to send and to receive data on a
wlan interface.
It has not to be connected to a wlan network as long as sender and
receiver are on the same channel. This enables you to get connection to
GNUnet where no internet access is possible, for example during
catastrophes or when censorship cuts you off from the internet.


@menu
* Requirements for the WLAN plugin::
* Configuration::
* Before starting GNUnet::
* Limitations and known bugs::
@end menu


@node Requirements for the WLAN plugin
@subsubsection Requirements for the WLAN plugin

@itemize @bullet

@item wlan network card with monitor support and packet injection
(see @uref{http://www.aircrack-ng.org/, aircrack-ng.org})

@item Linux kernel with mac80211 stack, introduced in 2.6.22, tested with
2.6.35 and 2.6.38

@item Wlantools to create the a monitor interface, tested with airmon-ng
of the aircrack-ng package
@end itemize

@node Configuration
@subsubsection Configuration

There are the following options for the wlan plugin (they should be like
this in your default config file, you only need to adjust them if the
values are incorrect for your system)

@example
# section for the wlan transport plugin
[transport-wlan]
# interface to use, more information in the
# "Before starting GNUnet" section of the handbook.
INTERFACE = mon0
# testmode for developers:
# 0 use wlan interface,
#1 or 2 use loopback driver for tests 1 = server, 2 = client
TESTMODE = 0
@end example

@node Before starting GNUnet
@subsubsection Before starting GNUnet

Before starting GNUnet, you have to make sure that your wlan interface is
in monitor mode.
One way to put the wlan interface into monitor mode (if your interface
name is wlan0) is by executing:

@example
sudo airmon-ng start wlan0
@end example

@noindent
Here is an example what the result should look like:

@example
Interface Chipset Driver
wlan0 Intel 4965 a/b/g/n iwl4965 - [phy0]
(monitor mode enabled on mon0)
@end example

@noindent
The monitor interface is mon0 is the one that you have to put into the
configuration file.

@node Limitations and known bugs
@subsubsection Limitations and known bugs

Wlan speed is at the maximum of 1 Mbit/s because support for choosing the
wlan speed with packet injection was removed in newer kernels.
Please pester the kernel developers about fixing this.

The interface channel depends on the wlan network that the card is
connected to. If no connection has been made since the start of the
computer, it is usually the first channel of the card.
Peers will only find each other and communicate if they are on the same
channel. Channels must be set manually, e.g. by using:

@example
iwconfig wlan0 channel 1
@end example

@node Configuring HTTP(S) reverse proxy functionality using Apache or nginx
@subsection Configuring HTTP(S) reverse proxy functionality using Apache or nginx

The HTTP plugin supports data transfer using reverse proxies. A reverse
proxy forwards the HTTP request he receives with a certain URL to another
webserver, here a GNUnet peer.

So if you have a running Apache or nginx webserver you can configure it to
be a GNUnet reverse proxy. Especially if you have a well-known website
this improves censorship resistance since it looks as normal surfing
behaviour.

To do so, you have to do two things:

@itemize @bullet
@item Configure your webserver to forward the GNUnet HTTP traffic
@item Configure your GNUnet peer to announce the respective address
@end itemize

As an example we want to use GNUnet peer running:

@itemize @bullet

@item HTTP server plugin on @code{gnunet.foo.org:1080}

@item HTTPS server plugin on @code{gnunet.foo.org:4433}

@item A apache or nginx webserver on
@uref{http://www.foo.org/, http://www.foo.org:80/}

@item A apache or nginx webserver on https://www.foo.org:443/
@end itemize

And we want the webserver to accept GNUnet traffic under
@code{http://www.foo.org/bar/}. The required steps are described here:

@menu
* Reverse Proxy - Configure your Apache2 HTTP webserver::
* Reverse Proxy - Configure your Apache2 HTTPS webserver::
* Reverse Proxy - Configure your nginx HTTPS webserver::
* Reverse Proxy - Configure your nginx HTTP webserver::
* Reverse Proxy - Configure your GNUnet peer::
@end menu

@node Reverse Proxy - Configure your Apache2 HTTP webserver
@subsubsection Reverse Proxy - Configure your Apache2 HTTP webserver

First of all you need mod_proxy installed.

Edit your webserver configuration. Edit
@code{/etc/apache2/apache2.conf} or the site-specific configuration file.

In the respective @code{server config},@code{virtual host} or
@code{directory} section add the following lines:

@example
ProxyTimeout 300
ProxyRequests Off
<Location /bar/ >
ProxyPass http://gnunet.foo.org:1080/
ProxyPassReverse http://gnunet.foo.org:1080/
</Location>
@end example

@node Reverse Proxy - Configure your Apache2 HTTPS webserver
@subsubsection Reverse Proxy - Configure your Apache2 HTTPS webserver

We assume that you already have an HTTPS server running, if not please
check how to configure a HTTPS host. An uncomplicated to use example
is the example configuration file for Apache2/HTTPD provided in
@file{apache2/sites-available/default-ssl}.

In the respective HTTPS @code{server config},@code{virtual host} or
@code{directory} section add the following lines:

@example
SSLProxyEngine On
ProxyTimeout 300
ProxyRequests Off
<Location /bar/ >
ProxyPass https://gnunet.foo.org:4433/
ProxyPassReverse https://gnunet.foo.org:4433/
</Location>
@end example

@noindent
More information about the apache mod_proxy configuration can be found
in the
@uref{http://httpd.apache.org/docs/2.2/mod/mod_proxy.html#proxypass, Apache documentation}.

@node Reverse Proxy - Configure your nginx HTTPS webserver
@subsubsection Reverse Proxy - Configure your nginx HTTPS webserver

Since nginx does not support chunked encoding, you first of all have to
install the @code{chunkin}
@uref{http://wiki.nginx.org/HttpChunkinModule, module}.

To enable chunkin add:

@example
chunkin on;
error_page 411 = @@my_411_error;
location @@my_411_error @{
chunkin_resume;
@}
@end example

@noindent
Edit your webserver configuration. Edit @file{/etc/nginx/nginx.conf} or
the site-specific configuration file.

In the @code{server} section add:

@example
location /bar/ @{
proxy_pass http://gnunet.foo.org:1080/;
proxy_buffering off;
proxy_connect_timeout 5; # more than http_server
proxy_read_timeout 350; # 60 default, 300s is GNUnet's idle timeout
proxy_http_version 1.1; # 1.0 default
proxy_next_upstream error timeout invalid_header http_500 http_503 http_502 http_504;
@}
@end example

@node Reverse Proxy - Configure your nginx HTTP webserver
@subsubsection Reverse Proxy - Configure your nginx HTTP webserver

Edit your webserver configuration. Edit @file{/etc/nginx/nginx.conf} or
the site-specific configuration file.

In the @code{server} section add:

@example
ssl_session_timeout 6m;
location /bar/
@{
proxy_pass https://gnunet.foo.org:4433/;
proxy_buffering off;
proxy_connect_timeout 5; # more than http_server
proxy_read_timeout 350; # 60 default, 300s is GNUnet's idle timeout
proxy_http_version 1.1; # 1.0 default
proxy_next_upstream error timeout invalid_header http_500 http_503 http_502 http_504;
@}
@end example

@node Reverse Proxy - Configure your GNUnet peer
@subsubsection Reverse Proxy - Configure your GNUnet peer

To have your GNUnet peer announce the address, you have to specify the
@code{EXTERNAL_HOSTNAME} option in the @code{[transport-http_server]}
section:

@example
[transport-http_server]
EXTERNAL_HOSTNAME = http://www.foo.org/bar/
@end example

@noindent
and/or @code{[transport-https_server]} section:

@example
[transport-https_server]
EXTERNAL_HOSTNAME = https://www.foo.org/bar/
@end example

@noindent
Now restart your webserver and your peer...

@node Blacklisting peers
@subsection Blacklisting peers

Transport service supports to deny connecting to a specific peer of to a
specific peer with a specific transport plugin using the blacklisting
component of transport service. With@ blacklisting it is possible to deny
connections to specific peers of@ to use a specific plugin to a specific
peer. Peers can be blacklisted using@ the configuration or a blacklist
client can be asked.

To blacklist peers using the configuration you have to add a section to
your configuration containing the peer id of the peer to blacklist and
the plugin@ if required.

Examples:

To blacklist connections to P565... on peer AG2P... using tcp add:

@c FIXME: This is too long and produces errors in the pdf.
@example
[transport-blacklist AG2PHES1BARB9IJCPAMJTFPVJ5V3A72S3F2A8SBUB8DAQ2V0O3V8G6G2JU56FHGFOHMQVKBSQFV98TCGTC3RJ1NINP82G0RC00N1520]
P565723JO1C2HSN6J29TAQ22MN6CI8HTMUU55T0FUQG4CMDGGEQ8UCNBKUMB94GC8R9G4FB2SF9LDOBAJ6AMINBP4JHHDD6L7VD801G = tcp
@end example

To blacklist connections to P565... on peer AG2P... using all plugins add:

@example
[transport-blacklist-AG2PHES1BARB9IJCPAMJTFPVJ5V3A72S3F2A8SBUB8DAQ2V0O3V8G6G2JU56FHGFOHMQVKBSQFV98TCGTC3RJ1NINP82G0RC00N1520]
P565723JO1C2HSN6J29TAQ22MN6CI8HTMUU55T0FUQG4CMDGGEQ8UCNBKUMB94GC8R9G4FB2SF9LDOBAJ6AMINBP4JHHDD6L7VD801G =
@end example

You can also add a blacklist client using the blacklist API. On a
blacklist check, blacklisting first checks internally if the peer is
blacklisted and if not, it asks the blacklisting clients. Clients are
asked if it is OK to connect to a peer ID, the plugin is omitted.

On blacklist check for (peer, plugin)

@itemize @bullet
@item Do we have a local blacklist entry for this peer and this plugin?
@item YES: disallow connection
@item Do we have a local blacklist entry for this peer and all plugins?
@item YES: disallow connection
@item Does one of the clients disallow?
@item YES: disallow connection
@end itemize

@node Configuration of the HTTP and HTTPS transport plugins
@subsection Configuration of the HTTP and HTTPS transport plugins

The client parts of the http and https transport plugins can be configured
to use a proxy to connect to the hostlist server. This functionality can
be configured in the configuration file directly or using the
gnunet-setup tool.

Both the HTTP and HTTPS clients support the following proxy types at
the moment:

@itemize @bullet
@item HTTP 1.1 proxy
@item SOCKS 4/4a/5/5 with hostname
@end itemize

In addition authentication at the proxy with username and password can be
configured.

To configure proxy support for the clients in the gnunet-setup tool,
select the "transport" tab and activate the respective plugin. Now you
can select the appropriate proxy type. The hostname or IP address
(including port if required) has to be entered in the "Proxy hostname"
textbox. If required, enter username and password in the "Proxy username"
and "Proxy password" boxes. Be aware that these information will be stored
in the configuration in plain text.

To configure these options directly in the configuration, you can
configure the following settings in the @code{[transport-http_client]}
and @code{[transport-https_client]} section of the configuration:

@example
# Type of proxy server,
# Valid values: HTTP, SOCKS4, SOCKS5, SOCKS4A, SOCKS5_HOSTNAME
# Default: HTTP
# PROXY_TYPE = HTTP

# Hostname or IP of proxy server
# PROXY =
# User name for proxy server
# PROXY_USERNAME =
# User password for proxy server
# PROXY_PASSWORD =
@end example

@node Configuring the GNU Name System
@subsection Configuring the GNU Name System

@menu
* Configuring system-wide DNS interception::
* Configuring the GNS nsswitch plugin::
@c * Configuring GNS on W32::
* GNS Proxy Setup::
* Setup of the GNS CA::
* Testing the GNS setup::
* Migrating existing DNS zones into GNS::
@end menu


@node Configuring system-wide DNS interception
@subsubsection Configuring system-wide DNS interception

Before you install GNUnet, make sure you have a user and group 'gnunet'
as well as an empty group 'gnunetdns'.

When using GNUnet with system-wide DNS interception, it is absolutely
necessary for all GNUnet service processes to be started by
@code{gnunet-service-arm} as user and group 'gnunet'. You also need to be
sure to run @code{make install} as root (or use the @code{sudo} option to
configure) to grant GNUnet sufficient privileges.

With this setup, all that is required for enabling system-wide DNS
interception is for some GNUnet component (VPN or GNS) to request it.
The @code{gnunet-service-dns} will then start helper programs that will
make the necessary changes to your firewall (@code{iptables}) rules.

Note that this will NOT work if your system sends out DNS traffic to a
link-local IPv6 address, as in this case GNUnet can intercept the traffic,
but not inject the responses from the link-local IPv6 address. Hence you
cannot use system-wide DNS interception in conjunction with link-local
IPv6-based DNS servers. If such a DNS server is used, it will bypass
GNUnet's DNS traffic interception.

Using the GNU Name System (GNS) requires two different configuration
steps.
First of all, GNS needs to be integrated with the operating system. Most
of this section is about the operating system level integration.

The remainder of this chapter will detail the various methods for
configuring the use of GNS with your operating system.

At this point in time you have different options depending on your OS:

@itemize @bullet
@item Use the gnunet-gns-proxy@*
This approach works for all operating systems and is likely the
easiest. However, it enables GNS only for browsers, not for other
applications that might be using DNS, such as SSH.  Still, using the
proxy is required for using HTTP with GNS and is thus recommended for
all users. To do this, you simply have to run the
@code{gnunet-gns-proxy-setup-ca} script as the user who will run the
browser (this will create a GNS certificate authority (CA) on your
system and import its key into your browser), then start
@code{gnunet-gns-proxy} and inform your browser to use the Socks5
proxy which @code{gnunet-gns-proxy} makes available by default on port
7777.
@item Use a nsswitch plugin (recommended on GNU systems)@*
This approach has the advantage of offering fully personalized
resolution even on multi-user systems. A potential disadvantage is
that some applications might be able to bypass GNS.
@item Use a W32 resolver plugin (recommended on W32)@*
This is currently the only option on W32 systems.
@item Use system-wide DNS packet interception@*
This approach is recommended for the GNUnet VPN. It can be used to
handle GNS at the same time; however, if you only use this method, you
will only get one root zone per machine (not so great for multi-user
systems).
@end itemize

You can combine system-wide DNS packet interception with the nsswitch
plugin.
The setup of the system-wide DNS interception is described here. All of
the other GNS-specific configuration steps are described in the following
sections.

@node Configuring the GNS nsswitch plugin
@subsubsection Configuring the GNS nsswitch plugin

The Name Service Switch (NSS) is a facility in Unix-like operating systems
(in most cases provided by the GNU C Library)
that provides a variety of sources for common configuration databases and
name resolution mechanisms.
A superuser (system administrator) usually configures the
operating system's name services using the file
@file{/etc/nsswitch.conf}.

GNS provides a NSS plugin to integrate GNS name resolution with the
operating system's name resolution process.
To use the GNS NSS plugin you have to either

@itemize @bullet
@item install GNUnet as root or
@item compile GNUnet with the @code{--with-sudo=yes} switch.
@end itemize

Name resolution is controlled by the @emph{hosts} section in the NSS
configuration. By default this section first performs a lookup in the
@file{/etc/hosts} file and then in DNS.
The nsswitch file should contain a line similar to:

@example
hosts: files dns [NOTFOUND=return] mdns4_minimal mdns4
@end example

@noindent
Here the GNS NSS plugin can be added to perform a GNS lookup before
performing a DNS lookup.
The GNS NSS plugin has to be added to the "hosts" section in
@file{/etc/nsswitch.conf} file before DNS related plugins:

@example
...
hosts: files gns [NOTFOUND=return] dns mdns4_minimal mdns4
...
@end example

@noindent
The @code{NOTFOUND=return} will ensure that if a @code{.gnu} name is not
found in GNS it will not be queried in DNS.

@c @node Configuring GNS on W32
@c @subsubsection Configuring GNS on W32

@c This document is a guide to configuring GNU Name System on W32-compatible
@c platforms.

@c After GNUnet is installed, run the w32nsp-install tool:

@c @example
@c w32nsp-install.exe libw32nsp-0.dll
@c @end example

@c @noindent
@c ('0' is the library version of W32 NSP; it might increase in the future,
@c change the invocation accordingly).

@c This will install GNS namespace provider into the system and allow other
@c applications to resolve names that end in '@strong{gnu}'
@c and '@strong{zkey}'. Note that namespace provider requires
@c gnunet-gns-helper-service-w32 to be running, as well as gns service
@c itself (and its usual dependencies).

@c Namespace provider is hardcoded to connect to @strong{127.0.0.1:5353},
@c and this is where gnunet-gns-helper-service-w32 should be listening to
@c (and is configured to listen to by default).

@c To uninstall the provider, run:

@c @example
@c w32nsp-uninstall.exe
@c @end example

@c @noindent
@c (uses provider GUID to uninstall it, does not need a dll name).

@c Note that while MSDN claims that other applications will only be able to
@c use the new namespace provider after re-starting, in reality they might
@c stat to use it without that. Conversely, they might stop using the
@c provider after it's been uninstalled, even if they were not re-started.
@c W32 will not permit namespace provider library to be deleted or
@c overwritten while the provider is installed, and while there is at least
@c one process still using it (even after it was uninstalled).

@node GNS Proxy Setup
@subsubsection GNS Proxy Setup

When using the GNU Name System (GNS) to browse the WWW, there are several
issues that can be solved by adding the GNS Proxy to your setup:

@itemize @bullet

@item If the target website does not support GNS, it might assume that it
is operating under some name in the legacy DNS system (such as
example.com). It may then attempt to set cookies for that domain, and the
web server might expect a @code{Host: example.com} header in the request
from your browser.
However, your browser might be using @code{example.gnu} for the
@code{Host} header and might only accept (and send) cookies for
@code{example.gnu}. The GNS Proxy will perform the necessary translations
of the hostnames for cookies and HTTP headers (using the LEHO record for
the target domain as the desired substitute).

@item If using HTTPS, the target site might include an SSL certificate
which is either only valid for the LEHO domain or might match a TLSA
record in GNS. However, your browser would expect a valid certificate for
@code{example.gnu}, not for some legacy domain name. The proxy will
validate the certificate (either against LEHO or TLSA) and then
on-the-fly produce a valid certificate for the exchange, signed by your
own CA. Assuming you installed the CA of your proxy in your browser's
certificate authority list, your browser will then trust the
HTTPS/SSL/TLS connection, as the hostname mismatch is hidden by the proxy.

@item Finally, the proxy will in the future indicate to the server that it
speaks GNS, which will enable server operators to deliver GNS-enabled web
sites to your browser (and continue to deliver legacy links to legacy
browsers)
@end itemize

@node Setup of the GNS CA
@subsubsection Setup of the GNS CA

First you need to create a CA certificate that the proxy can use.
To do so use the provided script gnunet-gns-proxy-ca:

@example
$ gnunet-gns-proxy-setup-ca
@end example

@noindent
This will create a personal certification authority for you and add this
authority to the firefox and chrome database. The proxy will use the this
CA certificate to generate @code{*.gnu} client certificates on the fly.

Note that the proxy uses libcurl. Make sure your version of libcurl uses
GnuTLS and NOT OpenSSL. The proxy will @b{not} work with libcurl compiled
against OpenSSL.

You can check the configuration your libcurl was build with by
running:

@example
curl --version
@end example

the output will look like this (without the linebreaks):

@example
gnurl --version
curl 7.56.0 (x86_64-unknown-linux-gnu) libcurl/7.56.0 \
GnuTLS/3.5.13 zlib/1.2.11 libidn2/2.0.4
Release-Date: 2017-10-08
Protocols: http https
Features: AsynchDNS IDN IPv6 Largefile NTLM SSL libz \
TLS-SRP UnixSockets HTTPS-proxy
@end example

@node Testing the GNS setup
@subsubsection Testing the GNS setup

Now for testing purposes we can create some records in our zone to test
the SSL functionality of the proxy:

@example
$ gnunet-identity -C test
$ gnunet-namestore -a -e "1 d" -n "homepage" \
  -t A -V 131.159.74.67 -z test
$ gnunet-namestore -a -e "1 d" -n "homepage" \
  -t LEHO -V "gnunet.org" -z test
@end example

@noindent
At this point we can start the proxy. Simply execute

@example
$ gnunet-arm -i gns-proxy
@end example

To run the proxy at all times in the future, you should
change your configuration as follows:

@example
$ gnunet-config -s gns-proxy -o AUTOSTART -V YES
@end example

@noindent
Configure your browser to use this SOCKSv5 proxy using
@code{localhost} on port 7777.
If you use @command{Firefox} (or one of its derivatives/forks such as
Icecat) you also have to go to @code{about:config} and set the key
@code{network.proxy.socks_remote_dns} to @code{true}.

When you visit @code{https://homepage.test/}, you should get to the
@code{https://gnunet.org/} frontpage and the browser (with the correctly
configured proxy) should give you a valid SSL certificate for
@code{homepage.gnu} and no warnings. It should look like this:

@c FIXME: Image does not exist, create it or save it from Drupal?
@c @image{images/gnunethpgns.png,5in,, picture of homepage.gnu in Webbrowser}

@node Migrating existing DNS zones into GNS
@subsubsection Migrating existing DNS zones into GNS

To migrate an existing zone into GNS use the Ascension tool.

Ascension transfers entire zones into GNS by doing incremental zone transfers
and then adding the records to GNS.

Compared to the gnunet-zoneimport tool it strictly uses AXFR or IXFR depending
on whether or not there exists a SOA record for the zone. If that is the case it
will take the serial as a reference point and request the zone. The server will
either answer the IXFR request with a correct incremental zone or with the
entire zone, which depends on the server configuration.

You can find the source code here: @code{https://git.gnunet.org/ascension.git/}

The software can be installed into a Python virtual environment like this:
@example
$ python3 -m venv .venv
$ source .venv/bin/activate
$ python3 setup.py install
@end example

Or installed globally like this:
@example
$ sudo python3 setup.py install
@end example

Pip will then install all the necessary requirements that are needed to run
Ascension. For development purposes a virtual environment should suffice.
Keeping a virtual environment helps with keeping things tidy and prevents
breaking of Ascension through a future Python update.

The advantage of using a virtual environment is, that all the dependencies can
be installed separately in different versions without touching your systems
Python installation and its dependencies.

Another way to install Ascension on Debian is to install the python3-ascension
package. It can be found within the above mentioned Ascension git repository.
This also adds a system user called ascension and runs a GNUnet peer in the
background. Please note: This only works if a recent version of GNUnet is
installed on your system. The version number of Ascension is chosen according
to the required feature level of GNUnet: Ascension 0.11.5 is only
compatible with GNUnet 0.11.5 or later and so on.
As Debian's packages for GNUnet are outdated even in experimental,
you will need to install GNUnet manually
@xref{Installing GNUnet}.

Please check @xref{Migrating an existing DNS zone into GNS}, for usage manual
of the tool.

@node Configuring the GNUnet VPN
@subsection Configuring the GNUnet VPN

@menu
* IPv4 address for interface::
* IPv6 address for interface::
* Configuring the GNUnet VPN DNS::
* Configuring the GNUnet VPN Exit Service::
* IP Address of external DNS resolver::
* IPv4 address for Exit interface::
* IPv6 address for Exit interface::
@end menu

Before configuring the GNUnet VPN, please make sure that system-wide DNS
interception is configured properly as described in the section on the
GNUnet DNS setup. @pxref{Configuring the GNU Name System},
if you haven't done so already.

The default options for the GNUnet VPN are usually sufficient to use
GNUnet as a Layer 2 for your Internet connection.
However, what you always have to specify is which IP protocol you want
to tunnel: IPv4, IPv6 or both.
Furthermore, if you tunnel both, you most likely should also tunnel
all of your DNS requests.
You theoretically can tunnel "only" your DNS traffic, but that usually
makes little sense.

The other options as shown on the gnunet-setup tool are:

@node IPv4 address for interface
@subsubsection IPv4 address for interface

This is the IPv4 address the VPN interface will get. You should pick a
'private' IPv4 network that is not yet in use for you system. For example,
if you use @code{10.0.0.1/255.255.0.0} already, you might use
@code{10.1.0.1/255.255.0.0}.
If you use @code{10.0.0.1/255.0.0.0} already, then you might use
@code{192.168.0.1/255.255.0.0}.
If your system is not in a private IP-network, using any of the above will
work fine.
You should try to make the mask of the address big enough
(@code{255.255.0.0} or, even better, @code{255.0.0.0}) to allow more
mappings of remote IP Addresses into this range.
However, even a @code{255.255.255.0} mask will suffice for most users.

@node IPv6 address for interface
@subsubsection IPv6 address for interface

The IPv6 address the VPN interface will get. Here you can specify any
non-link-local address (the address should not begin with @code{fe80:}).
A subnet Unique Local Unicast (@code{fd00::/8} prefix) that you are
currently not using would be a good choice.

@node Configuring the GNUnet VPN DNS
@subsubsection Configuring the GNUnet VPN DNS

To resolve names for remote nodes, activate the DNS exit option.

@node Configuring the GNUnet VPN Exit Service
@subsubsection Configuring the GNUnet VPN Exit Service

If you want to allow other users to share your Internet connection (yes,
this may be dangerous, just as running a Tor exit node) or want to
provide access to services on your host (this should be less dangerous,
as long as those services are secure), you have to enable the GNUnet exit
daemon.

You then get to specify which exit functions you want to provide. By
enabling the exit daemon, you will always automatically provide exit
functions for manually configured local services (this component of the
system is under
development and not documented further at this time). As for those
services you explicitly specify the target IP address and port, there is
no significant security risk in doing so.

Furthermore, you can serve as a DNS, IPv4 or IPv6 exit to the Internet.
Being a DNS exit is usually pretty harmless. However, enabling IPv4 or
IPv6-exit without further precautions may enable adversaries to access
your local network, send spam, attack other systems from your Internet
connection and do other mischiefs that will appear to come from your
machine. This may or may not get you into legal trouble.
If you want to allow IPv4 or IPv6-exit functionality, you should strongly
consider adding additional firewall rules manually to protect your local
network and to restrict outgoing TCP traffic (e.g. by not allowing access
to port 25). While we plan to improve exit-filtering in the future,
you're currently on your own here.
Essentially, be prepared for any kind of IP-traffic to exit the respective
TUN interface (and GNUnet will enable IP-forwarding and NAT for the
interface automatically).

Additional configuration options of the exit as shown by the gnunet-setup
tool are:

@node IP Address of external DNS resolver
@subsubsection IP Address of external DNS resolver

If DNS traffic is to exit your machine, it will be send to this DNS
resolver. You can specify an IPv4 or IPv6 address.

@node IPv4 address for Exit interface
@subsubsection IPv4 address for Exit interface

This is the IPv4 address the Interface will get. Make the mask of the
address big enough (255.255.0.0 or, even better, 255.0.0.0) to allow more
mappings of IP addresses into this range. As for the VPN interface, any
unused, private IPv4 address range will do.

@node IPv6 address for Exit interface
@subsubsection IPv6 address for Exit interface

The public IPv6 address the interface will get. If your kernel is not a
very recent kernel and you are willing to manually enable IPv6-NAT, the
IPv6 address you specify here must be a globally routed IPv6 address of
your host.

Suppose your host has the address @code{2001:4ca0::1234/64}, then
using @code{2001:4ca0::1:0/112} would be fine (keep the first 64 bits,
then change at least one bit in the range before the bitmask, in the
example above we changed bit 111 from 0 to 1).

You may also have to configure your router to route traffic for the entire
subnet (@code{2001:4ca0::1:0/112} for example) through your computer (this
should be automatic with IPv6, but obviously anything can be
disabled).

@node Bandwidth Configuration
@subsection Bandwidth Configuration

You can specify how many bandwidth GNUnet is allowed to use to receive
and send data. This is important for users with limited bandwidth or
traffic volume.

@node Configuring NAT
@subsection Configuring NAT

Most hosts today do not have a normal global IP address but instead are
behind a router performing Network Address Translation (NAT) which assigns
each host in the local network a private IP address.
As a result, these machines cannot trivially receive inbound connections
from the Internet. GNUnet supports NAT traversal to enable these machines
to receive incoming connections from other peers despite their
limitations.

In an ideal world, you can press the "Attempt automatic configuration"
button in gnunet-setup to automatically configure your peer correctly.
Alternatively, your distribution might have already triggered this
automatic configuration during the installation process.
However, automatic configuration can fail to determine the optimal
settings, resulting in your peer either not receiving as many connections
as possible, or in the worst case it not connecting to the network at all.

To manually configure the peer, you need to know a few things about your
network setup. First, determine if you are behind a NAT in the first
place.
This is always the case if your IP address starts with "10.*" or
"192.168.*". Next, if you have control over your NAT router, you may
choose to manually configure it to allow GNUnet traffic to your host.
If you have configured your NAT to forward traffic on ports 2086 (and
possibly 1080) to your host, you can check the "NAT ports have been opened
manually" option, which corresponds to the "PUNCHED_NAT" option in the
configuration file. If you did not punch your NAT box, it may still be
configured to support UPnP, which allows GNUnet to automatically
configure it. In that case, you need to install the "upnpc" command,
enable UPnP (or PMP) on your NAT box and set the "Enable NAT traversal
via UPnP or PMP" option (corresponding to "ENABLE_UPNP" in the
configuration file).

Some NAT boxes can be traversed using the autonomous NAT traversal method.
This requires certain GNUnet components to be installed with "SUID"
privileges on your system (so if you're installing on a system you do
not have administrative rights to, this will not work).
If you installed as 'root', you can enable autonomous NAT traversal by
checking the "Enable NAT traversal using ICMP method".
The ICMP method requires a way to determine your NAT's external (global)
IP address. This can be done using either UPnP, DynDNS, or by manual
configuration. If you have a DynDNS name or know your external IP address,
you should enter that name under "External (public) IPv4 address" (which
corresponds to the "EXTERNAL_ADDRESS" option in the configuration file).
If you leave the option empty, GNUnet will try to determine your external
IP address automatically (which may fail, in which case autonomous
NAT traversal will then not work).

Finally, if you yourself are not behind NAT but want to be able to
connect to NATed peers using autonomous NAT traversal, you need to check
the "Enable connecting to NATed peers using ICMP method" box.


@node Peer configuration for distributors (e.g. Operating Systems)
@subsection Peer configuration for distributors (e.g. Operating Systems)

The "GNUNET_DATA_HOME" in "[PATHS]" in @file{/etc/gnunet.conf} should be
manually set to "/var/lib/gnunet/data/" as the default
"~/.local/share/gnunet/" is probably not that appropriate in this case.
Similarly, distributors may consider pointing "GNUNET_RUNTIME_DIR" to
"/var/run/gnunet/" and "GNUNET_HOME" to "/var/lib/gnunet/". Also, should a
distributor decide to override system defaults, all of these changes
should be done in a custom @file{/etc/gnunet.conf} and not in the files
in the @file{config.d/} directory.

Given the proposed access permissions, the "gnunet-setup" tool must be
run as use "gnunet" (and with option "-c /etc/gnunet.conf" so that it
modifies the system configuration). As always, gnunet-setup should be run
after the GNUnet peer was stopped using "gnunet-arm -e". Distributors
might want to include a wrapper for gnunet-setup that allows the
desktop-user to "sudo" (e.g. using gtksudo) to the "gnunet" user account
and then runs "gnunet-arm -e", "gnunet-setup" and "gnunet-arm -s" in
sequence.

@node Config Leftovers
@section Config Leftovers

This section describes how to start a GNUnet peer. It assumes that you
have already compiled and installed GNUnet and its' dependencies.
Before you start a GNUnet peer, you may want to create a configuration
file using gnunet-setup (but you do not have to).
Sane defaults should exist in your
@file{$GNUNET_PREFIX/share/gnunet/config.d/} directory, so in practice
you could simply start without any configuration. If you want to
configure your peer later, you need to stop it before invoking the
@code{gnunet-setup} tool to customize further and to test your
configuration (@code{gnunet-setup} has built-in test functions).

The most important option you might have to still set by hand is in
[PATHS]. Here, you use the option "GNUNET_HOME" to specify the path where
GNUnet should store its data.
It defaults to @code{$HOME/}, which again should work for most users.
Make sure that the directory specified as GNUNET_HOME is writable to
the user that you will use to run GNUnet (note that you can run frontends
using other users, GNUNET_HOME must only be accessible to the user used to
run the background processes).

You will also need to make one central decision: should all of GNUnet be
run under your normal UID, or do you want distinguish between system-wide
(user-independent) GNUnet services and personal GNUnet services. The
multi-user setup is slightly more complicated, but also more secure and
generally recommended.

@menu
* The Single-User Setup::
* The Multi-User Setup::
* Killing GNUnet services::
* Access Control for GNUnet::
@end menu

@node The Single-User Setup
@subsection The Single-User Setup

For the single-user setup, you do not need to do anything special and can
just start the GNUnet background processes using @code{gnunet-arm}.
By default, GNUnet looks in @file{~/.config/gnunet.conf} for a
configuration (or @code{$XDG_CONFIG_HOME/gnunet.conf} if@
@code{$XDG_CONFIG_HOME} is defined). If your configuration lives
elsewhere, you need to pass the @code{-c FILENAME} option to all GNUnet
commands.

Assuming the configuration file is called @file{~/.config/gnunet.conf},
you start your peer using the @code{gnunet-arm} command (say as user
@code{gnunet}) using:

@example
gnunet-arm -c ~/.config/gnunet.conf -s
@end example

@noindent
The "-s" option here is for "start". The command should return almost
instantly. If you want to stop GNUnet, you can use:

@example
gnunet-arm -c ~/.config/gnunet.conf -e
@end example

@noindent
The "-e" option here is for "end".

Note that this will only start the basic peer, no actual applications
will be available.
If you want to start the file-sharing service, use (after starting
GNUnet):

@example
gnunet-arm -c ~/.config/gnunet.conf -i fs
@end example

@noindent
The "-i fs" option here is for "initialize" the "fs" (file-sharing)
application. You can also selectively kill only file-sharing support using

@example
gnunet-arm -c ~/.config/gnunet.conf -k fs
@end example

@noindent
Assuming that you want certain services (like file-sharing) to be always
automatically started whenever you start GNUnet, you can activate them by
setting "IMMEDIATE_START=YES" in the respective section of the configuration
file (for example, "[fs]"). Then GNUnet with file-sharing support would
be started whenever you@ enter:

@example
gnunet-arm -c ~/.config/gnunet.conf -s
@end example

@noindent
Alternatively, you can combine the two options:

@example
gnunet-arm -c ~/.config/gnunet.conf -s -i fs
@end example

@noindent
Using @code{gnunet-arm} is also the preferred method for initializing
GNUnet from @code{init}.

Finally, you should edit your @code{crontab} (using the @code{crontab}
command) and insert a line@

@example
@@reboot gnunet-arm -c ~/.config/gnunet.conf -s
@end example

to automatically start your peer whenever your system boots.

@node The Multi-User Setup
@subsection The Multi-User Setup

This requires you to create a user @code{gnunet} and an additional group
@code{gnunetdns}, prior to running @code{make install} during
installation.
Then, you create a configuration file @file{/etc/gnunet.conf} which should
contain the lines:@

@example
[arm]
START_SYSTEM_SERVICES = YES
START_USER_SERVICES = NO
@end example

@noindent
Then, perform the same steps to run GNUnet as in the per-user
configuration, except as user @code{gnunet} (including the
@code{crontab} installation).
You may also want to run @code{gnunet-setup} to configure your peer
(databases, etc.).
Make sure to pass @code{-c /etc/gnunet.conf} to all commands. If you
run @code{gnunet-setup} as user @code{gnunet}, you might need to change
permissions on @file{/etc/gnunet.conf} so that the @code{gnunet} user can
write to the file (during setup).

Afterwards, you need to perform another setup step for each normal user
account from which you want to access GNUnet. First, grant the normal user
(@code{$USER}) permission to the group gnunet:

@example
# adduser $USER gnunet
@end example

@noindent
Then, create a configuration file in @file{~/.config/gnunet.conf} for the
$USER with the lines:

@example
[arm]
START_SYSTEM_SERVICES = NO
START_USER_SERVICES = YES
@end example

@noindent
This will ensure that @code{gnunet-arm} when started by the normal user
will only run services that are per-user, and otherwise rely on the
system-wide services.
Note that the normal user may run gnunet-setup, but the
configuration would be ineffective as the system-wide services will use
@file{/etc/gnunet.conf} and ignore options set by individual users.

Again, each user should then start the peer using
@file{gnunet-arm -s} --- and strongly consider adding logic to start
the peer automatically to their crontab.

Afterwards, you should see two (or more, if you have more than one USER)
@code{gnunet-service-arm} processes running in your system.

@node Killing GNUnet services
@subsection Killing GNUnet services

It is not necessary to stop GNUnet services explicitly when shutting
down your computer.

It should be noted that manually killing "most" of the
@code{gnunet-service} processes is generally not a successful method for
stopping a peer (since @code{gnunet-service-arm} will instantly restart
them). The best way to explicitly stop a peer is using
@code{gnunet-arm -e}; note that the per-user services may need to be
terminated before the system-wide services will terminate normally.

@node Access Control for GNUnet
@subsection Access Control for GNUnet

This chapter documents how we plan to make access control work within the
GNUnet system for a typical peer. It should be read as a best-practice
installation guide for advanced users and builders of binary
distributions. The recommendations in this guide apply to POSIX-systems
with full support for UNIX domain sockets only.

Note that this is an advanced topic. The discussion presumes a very good
understanding of users, groups and file permissions. Normal users on
hosts with just a single user can just install GNUnet under their own
account (and possibly allow the installer to use SUDO to grant additional
permissions for special GNUnet tools that need additional rights).
The discussion below largely applies to installations where multiple users
share a system and to installations where the best possible security is
paramount.

A typical GNUnet system consists of components that fall into four
categories:

@table @asis

@item User interfaces
User interfaces are not security sensitive and are supposed to be run and
used by normal system users.
The GTK GUIs and most command-line programs fall into this category.
Some command-line tools (like gnunet-transport) should be excluded as they
offer low-level access that normal users should not need.
@item System services and support tools
System services should always run and offer services that can then be
accessed by the normal users.
System services do not require special permissions, but as they are not
specific to a particular user, they probably should not run as a
particular user. Also, there should typically only be one GNUnet peer per
host. System services include the gnunet-service and gnunet-daemon
programs; support tools include command-line programs such as gnunet-arm.
@item Privileged helpers
Some GNUnet components require root rights to open raw sockets or perform
other special operations. These gnunet-helper binaries are typically
installed SUID and run from services or daemons.
@item Critical services
Some GNUnet services (such as the DNS service) can manipulate the service
in deep and possibly highly security sensitive ways. For example, the DNS
service can be used to intercept and alter any DNS query originating from
the local machine. Access to the APIs of these critical services and their
privileged helpers must be tightly controlled.
@end table

@c FIXME: The titles of these chapters are too long in the index.

@menu
* Recommendation - Disable access to services via TCP::
* Recommendation - Run most services as system user "gnunet"::
* Recommendation - Control access to services using group "gnunet"::
* Recommendation - Limit access to certain SUID binaries by group "gnunet"::
* Recommendation - Limit access to critical gnunet-helper-dns to group "gnunetdns"::
* Differences between "make install" and these recommendations::
@end menu

@node Recommendation - Disable access to services via TCP
@subsubsection Recommendation - Disable access to services via TCP

GNUnet services allow two types of access: via TCP socket or via UNIX
domain socket.
If the service is available via TCP, access control can only be
implemented by restricting connections to a particular range of IP
addresses.
This is acceptable for non-critical services that are supposed to be
available to all users on the local system or local network.
However, as TCP is generally less efficient and it is rarely the case
that a single GNUnet peer is supposed to serve an entire local network,
the default configuration should disable TCP access to all GNUnet
services on systems with support for UNIX domain sockets.
Since GNUnet 0.9.2, configuration files with TCP access disabled should be
generated by default. Users can re-enable TCP access to particular
services simply by specifying a non-zero port number in the section of
the respective service.


@node Recommendation - Run most services as system user "gnunet"
@subsubsection Recommendation - Run most services as system user "gnunet"

GNUnet's main services should be run as a separate user "gnunet" in a
special group "gnunet".
The user "gnunet" should start the peer using "gnunet-arm -s" during
system startup. The home directory for this user should be
@file{/var/lib/gnunet} and the configuration file should be
@file{/etc/gnunet.conf}.
Only the @code{gnunet} user should have the right to access
@file{/var/lib/gnunet} (@emph{mode: 700}).

@node Recommendation - Control access to services using group "gnunet"
@subsubsection Recommendation - Control access to services using group "gnunet"

Users that should be allowed to use the GNUnet peer should be added to the
group "gnunet". Using GNUnet's access control mechanism for UNIX domain
sockets, those services that are considered useful to ordinary users
should be made available by setting "UNIX_MATCH_GID=YES" for those
services.
Again, as shipped, GNUnet provides reasonable defaults.
Permissions to access the transport and core subsystems might additionally
be granted without necessarily causing security concerns.
Some services, such as DNS, must NOT be made accessible to the "gnunet"
group (and should thus only be accessible to the "gnunet" user and
services running with this UID).

@node Recommendation - Limit access to certain SUID binaries by group "gnunet"
@subsubsection Recommendation - Limit access to certain SUID binaries by group "gnunet"

Most of GNUnet's SUID binaries should be safe even if executed by normal
users. However, it is possible to reduce the risk a little bit more by
making these binaries owned by the group "gnunet" and restricting their
execution to user of the group "gnunet" as well (4750).

@node Recommendation - Limit access to critical gnunet-helper-dns to group "gnunetdns"
@subsubsection Recommendation - Limit access to critical gnunet-helper-dns to group "gnunetdns"

A special group "gnunetdns" should be created for controlling access to
the "gnunet-helper-dns".
The binary should then be owned by root and be in group "gnunetdns" and
be installed SUID and only be group-executable (2750).
@b{Note that the group "gnunetdns" should have no users in it at all,
ever.}
The "gnunet-service-dns" program should be executed by user "gnunet" (via
gnunet-service-arm) with the binary owned by the user "root" and the group
"gnunetdns" and be SGID (2700). This way, @strong{only}
"gnunet-service-dns" can change its group to "gnunetdns" and execute the
helper, and the helper can then run as root (as per SUID).
Access to the API offered by "gnunet-service-dns" is in turn restricted
to the user "gnunet" (not the group!), which means that only
"benign" services can manipulate DNS queries using "gnunet-service-dns".

@node Differences between "make install" and these recommendations
@subsubsection Differences between "make install" and these recommendations

The current build system does not set all permissions automatically based
on the recommendations above. In particular, it does not use the group
"gnunet" at all (so setting gnunet-helpers other than the
gnunet-helper-dns to be owned by group "gnunet" must be done manually).
Furthermore, 'make install' will silently fail to set the DNS binaries to
be owned by group "gnunetdns" unless that group already exists (!).
An alternative name for the "gnunetdns" group can be specified using the
@code{--with-gnunetdns=GRPNAME} configure option.