.Dd $Mdocdate$
.Dt SNAC 8
.Os
.Sh NAME
.Nm snac
.Nd snac administration
.Sh DESCRIPTION
The
.Nm
daemon processes messages from other servers in the Fediverse
using the ActivityPub protocol.
.Pp
This is the admin manual. For user operation, see
.Xr snac 1 .
For file and data formats, see
.Xr snac 5 .
.Ss Special cares about your snac you must know beforehand
.Nm
makes heavy use of hard links and link reference counts for its work, so
don't even think of using it on a filesystem that doesn't support this
feature. Most UNIX-like operating systems (Linux, the BSDs, the old DEC
Ultrix machine in your grandfather basement, probably MacOS) support hard
links on their native filesystems. Don't do fancy things like moving the
subdirectories to different filesystems. Also, if you move your
.Nm
installation to another server, do it with a tool that respect hard
link counts. Remember:
.Nm
is a very UNIXy program that loves hard links.
.Ss Building and Installation
A C compiler must be installed in the system, as well as the development
headers and libraries for OpenSSL (or compatible) and curl. To build
.Nm ,
run
.Bd -literal -offset indent
make
.Ed
.Pp
And, after that, run as root
.Bd -literal -offset indent
make install
.Ed
.Ss Data storage Initialization
Once
.Nm
is properly installed on the system, designate a directory where
the server and user data are to be stored. This directory
must not exist yet.
.Nm
must always be run as a regular user; you can create one for
it or use your own. To initialize the data storage, execute
.Bd -literal -offset indent
snac init $HOME/snac-data
.Ed
.Pp
A small set of questions will be asked regarding the installation,
specially the host name it will run under, the local network address
and port
.Nm
will listen to, the optional path prefix and possibly other things.
.Pp
You can launch the
.Nm
process by running
.Bd -literal -offset indent
snac httpd $HOME/snac-data
.Ed
.Pp
Use a web browser to connect to the specified address and port. You
should see a greeting page.
.Pp
Log messages are sent to the standard error stream. By default, only
relevant information is written there. You can increase the debugging
level by editing the 'dbglevel' field in the
.Pa server.json
file or by setting a numeric value between 0 and 3 to the DEBUG
environment variable, see below.
.Pp
If you operate a Linux systemd-enabled system, OpenBSD or FreeBSD, there are
startup scripts and configuration data in the
.Pa examples
directory.
For other operating systems, please read the appropriate documentation
on how to install a daemon as a non-root service.
.Ss Upgrading to a new version
Sometimes, the data storage disk layout changes between versions. If there
is such a change,
.Nm
will refuse to run and require an upgrade. Do this by running
.Bd -literal -offset indent
snac upgrade $HOME/snac-data
.Ed
.Pp
Take special care to execute this upgrade operation without any
.Nm
processes serving on the same folder. You can break everything. I know
this because Tyler knows this.
.Pp
.Ss Server Setup
.Pp
An http server with TLS and proxying support must already be
installed and configured.
.Nm
runs as a daemon and listens on a TCP/IP socket, preferrably
on a local interface. It can serve the full domain or only
a directory. The http server must be configured to route to the
.Nm
socket all related traffic and also the webfinger standard
address. The Host header must be propagated.
See the examples below.
.Ss Adding Users
.Pp
Users must be created from the command line.
You can do it by running
.Bd -literal -offset indent
snac adduser $HOME/snac-data
.Ed
.Pp
All needed data will be prompted for. There is no artificial limit
on the number of users that can be created.
.Ss Customization
The
.Pa server.json
configuration file allows some behaviour tuning:
.Bl -tag -width tenletters
.It Ic host
The host name.
.It Ic prefix
The URL path prefix.
.It Ic address
The listen network address.
.It Ic port
The listen network port.
.It Ic dbglevel
The debug level. An integer value, being 0 the less verbose (the default).
.It Ic layout
The disk storage layout version. Never touch this.
.It Ic queue_retry_max
Messages sent out are stored in a queue. If the posting of a messages fails,
it's re-enqueued for later. This integer configures the maximum count of
times the sending will be retried.
.It Ic queue_retry_minutes
The number of minutes to wait before the failed posting of a message is
retried. This is not linear, but multipled by the number of retries
already done.
.It Ic max_timeline_entries
This is the maximum timeline entries shown in the web interface.
.It Ic timeline_purge_days
Entries in the timeline older that this number of days are purged.
If you don't want any timeline purging and enjoy your data drives
fill up with old crap and finally burst in flames, you can disable
purging by setting this to 0.
.It Ic local_purge_days
Same as before, but for the user-generated entries in the local timeline.
.It Ic cssurls
This is a list of URLs to CSS files that will be inserted, in this order,
in the HTML before the user CSS. Use these files to configure the global
site layout.
.It Ic disable_cache
If set to true, timeline caching is not done. This is only useful for
debugging purposes; don't enable it unless you know what do you want, as
it makes everything slower.
.It Ic disable_openbsd_security
If running under OpenBSD,
.Nm
makes use of the enhanced security functions
.Xr unveil 2
and
.Xr pledge 2 .
Setting this to true disables their usage. These functions limit severely
what an intruder can do in case of a security vulnerability, so only enable
this option if something is very broken.
.It Ic num_threads
By setting this value, you can specify the exact number of threads
.Nm
will use when processing connections. Values lesser than 4 will be ignored.
.It Ic disable_email_notifications
By setting this to true, no email notification will be sent for any user.
.It Ic disable_inbox_collection
By setting this to true, no inbox collection is done. Inbox collection helps
being discovered from remote instances, but also increases network traffic.
.It Ic http_headers
If you need to add more HTTP response headers for whatever reason, you can
fill this object with the required header/value pairs. For example, for enhanced
XSS security, you can set the "Content-Security-Policy" header to "script-src ;"
to be totally sure that no JavaScript is executed.
.It Ic show_instance_timeline
If this is set to true, the instance base URL will show a timeline with the latest
user posts instead of the default greeting static page. If other information
fields are set (see below), they are also shown.
.It Ic admin_email
The email address of the instance administrator (optional).
.It Ic admin_account
The user name of the instance administrator (optional).
.It Ic short_description
A textual short description about the instance (optional).
.It Ic fastcgi
If set to true,
.Nm
will use the FastCGI interface to communicate with the upper level
http server, that must be configured accordingly.
.It Ic disable_history
If set to true, history monthly snapshots are not served nor their links shown.
.It Ic shared_inboxes
This boolean value selects if shared inboxes are announced or not. Enabling
shared inboxes helps (somewhat) in optimizing incoming traffic for instances
with a large number of users.
.It Ic min_account_age
If this numeric value (in seconds) is set, any activity coming from an account
that was created more recently than that will be rejected. This may be used
to mitigate spam from automatically created accounts.
.El
.Pp
You must restart the server to make effective these changes.
.Pp
If a file named
.Pa greeting.html
is present in the server base directory, it will be returned whenever
the base URL of the server is requested. Fill it with whatever
information about the instance you want to supply to people
visiting the server, like sign up requirements, site policies
and such. The special %userlist% mark in the file will cause
the list of users in this instance to be inserted.
.Pp
Users can change a bit of information about themselves from the
web interface. See
.Xr snac 1
for details. Further, every user can have a private CSS file in their
.Pa static/style.css
that will be served instead of the server-wide one.
It's not modifiable from the web interface to avoid users
shooting themselves in the foot by destroying everything.
.Ss Custom Emojis
From version 2.51, support for customized Emojis in posts is available
(previously, they were hardcoded). Emojis are read from the
.Pa emojis.json
file in the instance base directory, as a JSON object of key / value
pairs (if this file does not exist, it will be created with
the predefined set). Each key in the object contains the text to be found (e.g.,
the :-) for a smiling face), and its associated value, the text string that
will replace it (in this example case, the HTML entity for the Unicode codepoint
for the smiley or the Emoji itself as text).
.Pp
Emoji values can also be URLs to image files; in this case, they will not be
substituted in the post content, but added to the 'tag' array as an ActivityPub
standard 'Emoji' object (it's recommendable that the Emoji key be enclosed in
colons for maximum compatilibity with other ActivityPub implementations, like
e.g. :happydoggo:). These images can be served from an external source or from the
.Pa static
directory of the instance admin.
.Pp
If you want to disable any Emoji substitution, change the file to contain
just an empty JSON object ({}).
.Ss SPAM Mitigation
There have been some SPAM attacks on the Fediverse and, as too many
instances and server implementations out there still allow automatic
account creation, it will only get worse.
.Nm
includes some (not very strong) tools for trying to survive the SPAM
flood that will eventually happen.
.Pp
The 
.Ic min_account_age
field in the main configuration file allows setting a minimum age (in
seconds) to consider too recently created accounts suspicious of being
a potential source of SPAM. This is a naïve assumption, because spammers
can create accounts, let them dormant for a while and then start to use
them. Also, some ActivityPub implementations don't even bother to return
a creation date for their accounts, so this is not very useful.
.Pp
From version 2.50, post content can be filtered out by regular expressions.
These weapons of mass destruction can be written into the
.Ic filter_reject.txt
file in the server base directory, one per line; if this file exists,
all posts' content will be matched (after being stripped of HTML tags)
against these regexes, one by one, and any match will make the post to
be rejected. If you don't know about regular expressions, don't use this
option (or learn about them in some tutorial, there are gazillions of
them out there), as you and your users may start missing posts. Also,
given that every regular expression implementation supports a different
set of features, consider reading the documentation about the one
implemented in your system.
.Ss ActivityPub Support
These are the following activities and objects that
.Nm
supports:
.Bl -tag -width tenletters
.It Vt Follow
Complete support, on input and output.
.It Vt Undo
For
.Vt Follow
objects, on input and output.
.It Vt Create
For
.Vt Note ,
.Vt Question
and
.Vt Page
objects, on input and output.
.It Vt Accept
For
.Vt Follow
objects, on input and output.
.It Vt Like
For
.Vt Note
objects, on input and output.
.It Vt Announce
For
.Vt Note
objects, on input and output.
.It Vt Update
For
.Vt Person ,
.Vt Note
and
.Vt Question
objects, on input and output.
.It Vt Delete
Supported for
.Vt Note
and
.Vt Tomsbtone
objects on input, and for
.Vt Note
objects on output.
.El
.Pp
The rest of activities and objects are dropped on input.
.Pp
There is partial support for
.Vt OrderedCollection
objects in the
.Pa /outbox 
(with the last 20 entries of the local timeline shown). No pagination
is supported. Intentionally, the
.Pa /followers
and
.Pa /following
paths return empty lists.
.Ss Migrating from Mastodon
User migration from different Fediverse instances is a pain in the ass
that has been implemented everywhere as a kludgy afterthought. There is
not much that can be done, other than importing the list of people you
follow to your new
.Nm
account.
.Pp
To do this, download the user's list of accounts being followed (in CSV
format) from the Mastodon web interface and execute this:
.Bd -literal -offset indent
awk -F, 'NR > 1 { print $1 }' /path/to/following_accounts.csv | \\
xargs -n 1 snac follow $SNAC_BASEDIR $SNAC_USER
.Ed
.Ss Instance blocking
Full instances can be blocked. This operation must be done from
the command-line tool. See
.Xr snac 1 .
.Ss Other Considerations
.Nm
stores all the messages it receives as JSON files, which are usually
bloated and filled with redundant information. Using a filesystem with
file compression enabled (like btrfs or zfs) will probably be a good
choice to store the
.Nm
data storage into.
.Sh ENVIRONMENT
.Bl -tag -width Ds
.It Ev DEBUG
Overrides the debugging level from the server 'dbglevel' configuration
variable. Set it to an integer value. The higher, the deeper in meaningless
verbiage you'll find yourself into.
.El
.Sh EXAMPLES
You want to install the
.Nm
Fediverse daemon in the host example.com, that is correctly configured
with a valid TLS certificate and running the nginx httpd server.
The service will be installed under the
.Pa fedi
location. Two users, walter and jessie, will be hosted in the system.
Their Fediverse presence addresses will be
.Lk https://example.com/fedi/walter
and
.Lk https://example.com/fedi/jesse ,
respectively. They will be known
in the Fediverse as @walter@example.com and @jesse@example.com. The
.Nm
daemon will run as the user snacusr in the system and listen to the
localhost:8001 network socket. All data will be stored in the
.Pa /home/snacusr/fedidata
directory.
.Pp
Log into the system as snacusr and execute:
.Bd -literal -offset indent
snac init /home/snacusr/fedidata
.Ed
.Pp
Answer "example.com" to the host name question, "/fedi" to the path
prefix question, "localhost" to the address and "8001" to the port.
.Pp
Create the users
.Bd -literal -offset indent
snac adduser /home/snacusr/fedidata walter
snac adduser /home/snacusr/fedidata jesse
.Ed
.Pp
Answer the questions with reasonable values.
.Pp
Execute the server:
.Bd -literal -offset indent
snac httpd /home/snacusr/fedidata
.Ed
.Pp
Edit the nginx configuration and add the following snippet to the
example.com server section:
.Bd -literal -offset indent
# nginx configuration example

# main web access point
location /fedi {
    proxy_pass http://localhost:8001;
    proxy_set_header Host $http_host;
}
# webfinger
location /.well-known/webfinger {
    proxy_pass http://localhost:8001;
    proxy_set_header Host $http_host;
}
# Mastodon API (entry points)
location /api/v1/ {
    proxy_pass http://localhost:8001;
    proxy_set_header Host $http_host;
}
location /api/v2/ {
    proxy_pass http://localhost:8001;
    proxy_set_header Host $http_host;
}
# Mastodon API (OAuth support)
location /oauth {
    proxy_pass http://localhost:8001;
    proxy_set_header Host $http_host;
}
# optional
location /.well-known/nodeinfo {
    proxy_pass http://localhost:8001;
    proxy_set_header Host $http_host;
}
# optional (needed by some Mastodon API clients)
location /.well-known/host-meta {
    proxy_pass http://localhost:8001;
    proxy_set_header Host $http_host;
}
.Ed
.Pp
Restart the nginx daemon and connect to
.Lk https://example.com/fedi/walter .
The empty, default screen will be shown. Enter the admin section with the
credentials defined for this user. Search people, start following
them, engage in arid discussions and generally enjoy the frustrating
experience of Social Media.
.Pp
This is an example of a similar configuration for the Apache2 web server:
.Bd -literal -offset indent
# apache2 configuration example

ProxyPreserveHost On

# Main web access point
<Location /fedi>
    ProxyPass http://127.0.0.1:8001/social
</Location>

# WebFinger
<Location /.well-known/webfinger>
    ProxyPass http://127.0.0.1:8001/.well-known/webfinger
</Location>

# Mastodon API (entry points)
<Location /api/v1/>
    ProxyPass http://127.0.0.1:8001/api/v1/
</Location>

<Location /api/v2/>
    ProxyPass http://127.0.0.1:8001/api/v2/
</Location>

# Mastodon API (OAuth support)
<Location /oauth>
    ProxyPass http://127.0.0.1:8001/oauth
</Location>

# NodeInfo (optional)
<Location /.well-known/nodeinfo>
    ProxyPass http://127.0.0.1:8001/.well-known/nodeinfo
</Location>

# host-meta (optional, needed for some Mastodon API clients)
<Location /.well-known/host-meta>
    ProxyPass http://127.0.0.1:8001/.well-known/host-meta
</Location>
.Ed
.Pp
Since version 2.43,
.Nm
supports communicating from / to the front end http server using the FastCGI
protocol. There is no special advantage in using this, only that some servers
allow for simpler configuration. For example, in the case of nginx, you can
replace the two 'proxy_pass' and 'proxy_set_header' lines in the example
above with just
.Bd -literal -offset indent
fastcgi_pass localhost:8001;
.Ed
.Pp
The only thing to change on
.Nm
size is to the set 'fastcgi' value to true in
.Pa server.json .
.Pp
Further, using the FastCGI interface allows a much simpler configuration
under OpenBSD's native httpd, given that it's natively implemented there
and you no longer need to configure the complicated relayd server. This is
an example:
.Bd -literal -offset indent
# OpenBSD httpd configuration example

# other server configuration
[...]

location "/fedi/*" {
    fastcgi socket tcp "127.0.0.1" 8001
}

location "/.well-known/webfinger" {
    fastcgi socket tcp "127.0.0.1" 8001
}

location "/oauth/*" {
    fastcgi socket tcp "127.0.0.1" 8001
}

location "/api/v1/*" {
    fastcgi socket tcp "127.0.0.1" 8001
}

location "/api/v2/*" {
    fastcgi socket tcp "127.0.0.1" 8001
}

location "/.well-known/nodeinfo" {
    fastcgi socket tcp "127.0.0.1" 8001
}

location "/.well-known/host-meta" {
    fastcgi socket tcp "127.0.0.1" 8001
}
.Ed
.Sh SEE ALSO
.Xr snac 1 ,
.Xr snac 5
.Sh AUTHORS
.An grunfink Lk https://comam.es/snac/grunfink @grunfink@comam.es
.Sh LICENSE
See the LICENSE file for details.
.Sh CAVEATS
JSON files are fragile when modified by hand. Take care.