Fossil

#!/bin/bash
BASEPORT=12345
FOSSIL=fossil
OLDPID=`pgrep -P 1 fossil`
PGARGS="-P 1"

if [ "$1" = "-f" ] ; then PGARGS= ; shift ; fi

if [ -n "$OLDPID" ]
then
    echo "Killing running Fossil server first..."
    pkill $PGARGS fossil

    for i in $(seq 30)
    do
        if [ -n "$(pgrep $PGARGS fossil)" ]
        then
            if [ $i -eq 1 ]
            then
                echo -n "Waiting for it to die..."
            else
                echo -n .
            fi
            sleep '0.1'
        else
            break
        fi
        echo
    done

    killall -9 fossil 2> /dev/null
fi

if [ -x ./fossil ]
then
    # We're running from a build tree, so use that version instead
    FOSSIL=./fossil
fi

function start_one() {
    bn=$1
    port=$(($BASEPORT + $2))
    url="$3"

    $FOSSIL server --localhost --port $port --scgi \
            --baseurl $url --errorlog ~/log/fossil/$bn-errors.log \
            ~/museum/$bn.fossil > ~/log/fossil/$bn-stdout.log &
    echo Fossil server running for $bn, PID $!, port $port.
}

start_one example 0 https://example.com/code
start_one foo     1 https://foo.net

<title>SSL and Fossil</title>

<h2>Using SSL with Fossil</h2>

If you are storing sensitive information in your repository, you should use SSL to encrypt all communications. This will protect the credentials used to access the server, as well preventing eavesdropping of the contents of your repository.







To host a repository with SSL, you need to use an web server which supports SSL in front of the Fossil server. You can host it using the CGI option or by proxying Fossil's built in HTTP server.
















Your fossil client must be built with SSL support. The <tt>configure</tt> script will attempt to find OpenSSL on your system, but if necessary, you can specify the location with the <tt>--with-openssl</tt> option. Type <tt>./configure --help</tt> for details.




Make sure the URL you clone from uses the <tt>https:</tt> scheme to ensure you're using SSL. If your server is configured to serve the repository from http as well as https, it's easy to accidentally use unencrypted HTTP if you forget the all important 's'.



























<h2>Certificates</h2>

To verify the identify of a server, SSL uses certificates. Fossil needs to know which certificates you trust.


If you are using a self-signed certificate, you'll be asked if you want to accept the certificate the first time you communicate with the server. Verify the certificate fingerprint is correct, then answer "always" to remember your decision.




If you are using a certificate signed by a certificate authority, you need to specify the certificates you trust with the <tt>ssl-ca-location</tt> setting. Set this globally with the <tt>-global</tt> option for convenience.







This should be set to the location of a file containing all the PEM encoded certificates you trust. You can obtain a certificate using a web browser, for example, Firefox, or just refer to your system's trusted CA roots which are usually stored somewhere in <tt>/etc</tt>.
















































<h2>Client side certificates</h2>


You can also use client side certificates to add an extra layer of authentication, over and above Fossil's built in user management. If you are particularly paranoid, you'll want to use this to remove the ability of anyone on the internet from making any request to Fossil. Without presenting a valid client side certificate, the web server won't invoke the fossil CGI handler.





Configure your server to request a client side certificate, and set up a certificate authority to sign your client certificates. For each person who needs to access the repository, create a private key and certificate signed with that CA.




The PEM encoded private key and certificate should be stored in a single file, simply by concatenating the key and certificate files. Specify the location of this file with the <tt>ssl-identity</tt> setting, or the <tt>--ssl-identity</tt> option to the <tt>clone</tt> command.





If you've password protected the private key, the password will be requested every time you connect to the server. This password is not stored by fossil, as doing so would defeat the purpose of having a password.




If you attempt to connect to a server which requests a client certificate, but don't provide one, fossil will show an error message which explains what to do to authenticate with the server.

<title>TLS and Fossil</title>

<h2>Using TLS-Encrypted Communications with Fossil</h2>

If you are storing sensitive information in a repository accessible over
a network whose integrity you do not fully trust, you should use TLS to
encrypt all communications with it. This is most true for repositories
accessed over the Internet, especially if they will be accessed from
edge networks you do not control, since that admits of various forms of
[https://en.wikipedia.org/wiki/Man-in-the-middle_attack|man in the
middle attack].


TLS protects the credentials used to access the server, prevents
eavesdropping, prevents in-flight data modification, prevents server
identify spoofing, and more.

There are two major aspects to this, both of which have to be addressed
in different ways. Those are the subjects of the next two major
sections.


<h2 id="client">Client-Side TLS Configuration</h2>

Fossil itself has built-in support for TLS on the client side only. That
is to say, you can build it against [https://www.openssl.org/|the
OpenSSL library], which will allow it to clone and sync with a remote
Fossil repository via <tt>https</tt> URIs.

The <tt>configure</tt> script will attempt to find OpenSSL on your
system automatically, but if necessary, you can specify the location
with the <tt>--with-openssl</tt> option. Type <tt>./configure
--help</tt> for details.


Even if the Fossil build system does manage to find a workable version
of OpenSSL, a common situation is that the platform version is outdated
in some key way, enough so that you do not want to use it with Fossil.
For example, the platform version of OpenSSL might not support any of
the [https://en.wikipedia.org/wiki/Cipher_suite|cipher suites] the
remote Fossil repository's HTTPS proxy is willing to offer, so that even
though both sides are speaking TLS/SSL, they can't come to an agreement
on the cryptography.
In such cases, you may want to link Fossil to a newer version of OpenSSL
than the one provided with your client operating system. You can do this
like so:

<pre>
    cd compat
    tar xf /path/to/openssl-*.tar.gz
    ln -fs openssl-x.y.z openssl
    ./config              # or, e.g. ./Configure darwin64-x86_64-cc
    make -j11
    cd ../..
    ./configure --with-openssl=tree
    make -j11
</pre>

That will get you a Fossil binary statically linked to this in-tree
version of OpenSSL.


<h3 id="certs">Certificates</h3>

To verify the identify of a server, TLS uses
[https://en.wikipedia.org/wiki/X.509#Certificates|X.509 certificates].

If you are using a self-signed certificate, you'll be asked if you want
to accept the certificate the first time you communicate with the
server. Verify the certificate fingerprint is correct, then answer
"always" to remember your decision.


If you are cloning from or syncing to Fossil servers that use a
certificate signed by a
[https://en.wikipedia.org/wiki/Certificate_authority|certificate
authority] (CA), Fossil needs to know which CAs you trust to sign those
certificates. Fossil relies on the OpenSSL library to have some way to
check a trusted list of CA signing keys.


There are two common ways this fails:

  #  <p>The OpenSSL library Fossil is linked to doesn't have a CA
     signing key set at all, so that it initially trusts no certificates
     at all.</p>
  #  <p>The OpenSSL library does have a CA cert set, but your Fossil server's
     TLS certificate was signed by a CA that isn't in that set.</p>

A common reason to fall into the second trap is that you're using
certificates signed by a local private CA, as often happens in large
enterprises.  You can solve this sort of problem by getting your local
CA's signing certificate in PEM format and pointing OpenSSL at it:

<pre>
     fossil set --global ssl-ca-location /path/to/local-ca.pem
</pre>

The use of <tt>--global</tt> with this option is common, since you may
have multiple reposotories served under certificates signed by that same
CA.

A common way to run into the broader first problem is that you're on
FreeBSD, which does not install a CA certificate set by default, even as
a dependency of the OpenSSL library.  If you're using a certificate
signed by one of the major public CAs, you can solve this by installing
the <tt>ca_root_nss</tt> package. That package contains the Mozilla NSS
certificate bundle, which gets installed in a location that OpenSSL
checks by default, so you don't need to change any Fossil settings.
(This is the same certificate set that ships with Firefox, by the way.)

The same sort of thing happens with the Windows build of OpenSSL, but
for a different core reason: Windows does ship with a stock CA
certificate set, but it's not in a format that OpenSSL understands how
to use.  Rather than try to find a way to convert the data format, you
may find it acceptable to use the same Mozilla NSS cert set.  I do not
know of a way to easily get this from Mozilla themselves, but I did find
a [https://curl.haxx.se/docs/caextract.html|third party source] for the
<tt>cacert.pem</tt> file. Install it somewhere on your system, then
point Fossil at it like so:

<pre>
     fossil set --global ssl-ca-location /path/to/cacert.pem
</pre>

Linux platforms tend to provide such a root cert store along with the
platform OpenSSL package, either built-in or as a hard dependency.


<h4>Client-Side Certificates</h4>

You can also use client side certificates to add an extra layer of
authentication, over and above Fossil's built in user management. If you
are particularly paranoid, you'll want to use this to remove the ability
of anyone on the internet from making any request to Fossil. Without
presenting a valid client side certificate, the web server won't invoke
the Fossil CGI handler.

Configure your server to request a client side certificate, and set up a
certificate authority to sign your client certificates. For each person
who needs to access the repository, create a private key and certificate
signed with that CA.

The PEM encoded private key and certificate should be stored in a single
file, simply by concatenating the key and certificate files. Specify the
location of this file with the <tt>ssl-identity</tt> setting, or the
<tt>--ssl-identity</tt> option to the <tt>clone</tt> command.

If you've password protected the private key, the password will be
requested every time you connect to the server. This password is not
stored by fossil, as doing so would defeat the purpose of having a
password.

If you attempt to connect to a server which requests a client
certificate, but don't provide one, fossil will show an error message
which explains what to do to authenticate with the server.


<h2 id="server">Server-Side TLS Configuration</h2>

Fossil's built-in HTTP server feature does not currently have a built-in
way to serve via HTTP over TLS, a.k.a. HTTPS, even when you've linked
Fossil to OpenSSL. To serve a Fossil repository via HTTPS, you must put
it behind some kind of HTTPS proxy.


<h3 id="stunnel">stunnel Alone</h3>

Conceptually, the simplest option is to [https://www.stunnel.org/|set up
stunnel].  A typical configuration is to run Fossil as an HTTP server
bound to localhost only, then export access to it via HTTPS with stunnel
encrypting access to Fossil instance hiding behind it.

The difficulty comes in configuring it, which really wants a guide that
no one has written for us yet. Until that appears, you'll have to read
the stunnel documentation and that of your TLS certificate provider to
work out how to set this up.


<h3 id="althttpd">stunnel + althttpd</h3>

The public SQLite and Fossil web sites can't just use stunnel + Fossil
because parts of the web site are static, served by
[https://www.sqlite.org/docsrc/file/misc/althttpd.c|a separate web
server called <tt>althttpd</tt>], written by the primary author of both
SQLite and Fossil. It is a lightweight HTTP-only web server. It handles
the static HTTP hits on <tt>sqlite.org</tt> and <tt>fossil-scm.org</tt>,
delegating HTTPS and dynamic Fossil content hits to stunnel and Fossil.

The only documentation for althttpd currently is in its header comment.
As is typical for drh software, althttpd is a single-file C program, so
that at worst, you just have to read its code to understand it.


<h3 id="nginx">nginx</h3>

If your needs are more complex than althttpd can handle or you'd prefer
to use only software available in your server operating system's package
repository, we recommend that you step up to [http://nginx.org/|nginx].
Setting this up is complex enough that we cover it [./tls-nginx.md|in a
separate document].


<h2 id="enforcing">Enforcing TLS Access</h2>

To use TLS encryption in cloning and syncing to a remote Fossil
repository, be sure to use the <tt>https:</tt> URI scheme in
<tt>clone</tt> and <tt>sync</tt> commands.  If your server is configured
to serve the repository via both HTTP and HTTPS, it's easy to
accidentally use unencrypted HTTP if you forget the all-important 's'.

There is a setting in the Fossil UI under Admin &rarr; Access called
"Redirect to HTTPS on the Login page."  This setting is not enabled by
default.  This setting does not automatically upgrade clones and syncs
done via the <tt>http</tt> URI scheme.  It only affects web UI access to
the Fossil repository.

<b id="rloop">WARNING:</b> Never enable this setting when running Fossil
behind an HTTPS proxy with Fossil running underneath it via HTTP or
SCGI.  This will cause an infinite redirect loop any time someone tries
to log into the web UI.  Fossil sees that it's being accessed via HTTP,
so it redirects the browser to an HTTPS equivalent URL, which causes the
client to hit the HTTPS front end proxy up again for access, which
causes Fossil to see that it's being accessed via HTTP, so it redirects
the client to...'round and 'round it goes until the web browser detects
it's in a redirect loop and gives up.

If you wish to enforce TLS-only access to a Fossil web server, it is
best done at the HTTPS front-end proxy layer, not by use of Fossil-level
settings like this one.  The [./tls-nginx.md|nginx TLS proxy guide]
shows one way to achieve this, for example.


<h2>Terminology Note</h2>

This document is called <tt>ssl.wiki</tt> for historical reasons. The
TLS protocol was originally called SSL, and it went through several
revisions before being replaced by TLS. Years before this writing, SSL
finally became entirely obsolete due to weaknesses in the protocol fixed
in the later TLS series of protocols.

Some people still use the term "SSL" when they actually mean "TLS," but
in the Fossil project, we always use "TLS" except when we must preserve
some sort of historical compatibility, as with this document's name in
order to avoid broken external URLs.  The Fossil TLS-related settings
also often use "`ssl`" in their names, for the same reason.

This series of protocols is also called "HTTPS" after the URI scheme
used to specify "HTTP over TLS."

# Proxying Fossil via HTTPS with nginx

One of the [many ways](./ssl.wiki) to provide TLS-encrypted HTTP access
(a.k.a. HTTPS) to Fossil is to run it behind a web proxy that supports
TLS. This document explains how to use the powerful [nginx web
server](http://nginx.org/) to do that.


## Benefits

This scheme is complicated, even with the benefit of this guide and
pre-built binary packages. Why should you put up with this complexity?
Because it gives many benefits that are difficult or impossible to get
with the less complicated options:

*   **Power** — nginx is one of the most powerful web servers in the
    world. The chance that you will run into a web serving wall that you
    can’t scale with nginx is very low.

    To give you some idea of the sort of thing you can readily
    accomplish with nginx, your author runs a single public web server
    that provides transparent name-based virtual hosting for four
    separate domains:

    *   One is entirely static, not involving any dynamic content or
        Fossil integration at all.

    *   Another is served almost entirely by Fossil, with a few select
        static content exceptions punched past Fossil, which are handled
        entirely via nginx.

    *   The other two domains are aliases for one another — e.g.
        `example.com` and `example.net` — with most of the content being
        static.  This pair of domains has three different Fossil repo
        proxies attached to various sections of the URI hierarchy.

    All of this is done with minimal configuration repetition between
    the site configurations.

*   **Integration** — Because nginx is so popular, it integrates with
many different technologies, and many other systems integrate with it in
turn.  This makes it great middleware, sitting between the outer web
world and interior site services like Fossil. It allows Fossil to
participate seamlessly as part of a larger web stack.

*   **Availability** — nginx is already in most operating system binary
package repositories, so you don’t need to go out of your way to get it.


## Fossil Remote Access Methods

Fossil provides four major ways to access a repository it’s serving
remotely, three of which you can use with nginx:

*   **HTTP** — Fossil has a built-in HTTP server: `fossil server`.
    While this method is efficient and it’s possible to use nginx to
    proxy access to another HTTP server, this option is overkill for our
    purposes.  nginx is itself a fully featured HTTP server, so we will
    choose in this guide not to make nginx reinterpret Fossil’s
    implementation of HTTP.

*   **CGI** — This method is simple but inefficient, because it launches
    a separate Fossil instance on every HTTP hit.
    
    Since Fossil is a relatively small self-contained program, and it’s
    designed to start up quickly, this method can work well in a
    surprisingly large number of cases.

    Nevertheless, we will avoid this option in this document because
    we’re already buying into a certain amount of complexity here in
    order to gain power.  There’s no sense in throwing away any of that
    hard-won performance on CGI overhead.

*   **SCGI** — The [SCGI protocol][scgi] provides the simplicity of CGI
    without its performance problems.

*   **SSH** — This method exists primarily to avoid the need for HTTPS
    in the first place.  There is probably a way to get nginx to proxy
    Fossil to HTTPS via SSH, but it would be pointlessly complicated.

SCGI it is, then.


# Installing

The first step is to install the pieces we’ll be working with.  This
varies on different operating systems, so to avoid overcomplicating this
guide, we’re going to assume you’re using Ubuntu Server 18.04 LTS, a
common Tier 1 offering for [virtual private servers][vps].

SSH into your server, then say:

       $ sudo apt install certbot fossil nginx

For other operating systems, simply visit [the front Certbot web
page][cb] and tell it what OS and web stack you’re using. Chances are
good that they’ve got a good guide for you already.


# Running Fossil in SCGI Mode

You presumably already have a working Fossil configuration on the public
server you’re trying to set up and are just following this guide to
replace HTTP service with HTTPS.

(You can adjust the advice in this guide to get both HTTP *and* HTTPS
service on the same site, but I strongly recommend that you do not do
that: the good excuses remaining for continuing to allow HTTP on public
web servers are running thin these days.)

I run my Fossil SCGI server instances with a variant of [the `fslsrv`
shell script](/file/tools/fslsrv) currently hosted in the Fossil source
code repository. You’ll want to download that and make a copy of it, so
you can customize it to your particular needs.

This script allows running multiple Fossil SCGI servers, one per
repository, each bound to a different high-numbered `localhost` port, so
that only nginx can see and proxy them out to the public.  The
“`example`” repo is on TCP port localhost:12345, and the “`foo`” repo is
on localhost:12346.

As written, the `fslsrv` script expects repositories to be stored in the
calling user’s home directory under `~/museum`, because where else do
you keep Fossils?

That home directory also needs to have a directory to hold log files,
`~/log/fossil/*.log`. Fossil doesn’t put out much logging, but when it
does, it’s better to have it captured than to need to re-create the
problem after the fact.

The use of `--baseurl` in this script lets us have each Fossil
repository mounted in a different location in the URL scheme.  Here, for
example, we’re saying that the “`example`” repository is hosted under
the `/code` URI on its domains, but that the “`foo`” repo is hosted at
the top level of its domain.  You’ll want to do something like the
former for a Fossil repo that’s just one piece of a larger site, but the
latter for a repo that is basically the whole point of the site.

This script’s automatic restart feature makes Fossil upgrades easy:

       $ cd ~/src/fossil/trunk ; fossil up ; make ; killall fossil ;
         sudo make install ; fslsrv

I’ve written that as a single long command because I keep it in the
history for my Fossil servers, so I can just run it again from history.
You could put it in a shell script instead.

The `killall fossil` step is needed only on OSes that refuse to let you
replace a running binary on disk.

As written, the `fslsrv` script assumes a Linux environment.  It expects
`/bin/bash` to exist, and it depends on non-POSIX tools like `pgrep`.
It shouldn’t be difficult to port to very different systems, like macOS
or the BSDs.


# Configuring Let’s Encrypt, the Easy Way

If your web serving needs are simple, [Certbot][cb] can configure nginx
for you and keep its certificates up to date.  The details are pretty
much as in the Certbot documentation for [nginx on Ubuntu 18.04 LTS
guide][cbnu], except that where they recommend that you use the
first-party Certbot packages, we’ve found that the ones that come with
Ubuntu work just fine.

The primary local configuration you need is to tell nginx how to proxy
certain URLs down to the Fossil instance you started above with the
`fslsrv` script:

      location / {
           include scgi_params;
           scgi_pass 127.0.0.1:12345;
           scgi_param HTTPS "on";
           scgi_param SCRIPT_NAME "";
      }

The TCP port number in that snippet is the key: it has to match the port
number generated by `fslsrv` from the base port number passed to the
`start_one` function.


# Configuring Let’s Encrypt, the Hard Way

If you’re finding that you can’t get certificates to be issued or
renewed using the Easy Way instructions, the problem is usually that
your nginx configuration is too complicated for Certbot’s `--nginx`
plugin to understand. It attempts to rewrite your nginx configuration
files on the fly to achieve the renewal, and if it doesn’t put its
directives in the right locations, the ACME verification steps can fail.

Your author’s configuration, glossed above, is complicated enough that
the current version of Certbot (0.28 at the time of this writing) can’t
cope with it.  That’s the primary motivation for me to write this guide:
I’m addressing the “me” years hence who needs to upgrade to Ubuntu 20.04
or 22.04 LTS and has forgotten all of this stuff. 😉


## Step 1: Shifting into Manual

The first thing to do is to turn off all of the Certbot automation,
because it’ll only get in our way.  First, disable the Certbot package’s
automatic background updater:

      $ sudo systemctl disable certbot.timer

Next, edit `/etc/letsencrypt/renewal/example.com.conf` to disable the
nginx plugins. You’re looking for two lines setting the “install” and
“auth” plugins to “nginx”.  You can comment them out or remove them
entirely.


## Step 2: Configuring nginx

On Ubuntu systems, at least, the primary user-level configuration file
is `/etc/nginx/sites-enabled/default`. For a configuration like I
described at the top of this article, I recommend that this file contain
only a list of include statements, one for each site that server hosts:

      include local/example
      include local/foo

Those files then each define one domain’s configuration.  Here,
`/etc/nginx/local/example` contains the configuration for
`*.example.com` and `*.example.net`; and `local/foo` contains the
configuration for `*.foo.net`.

Here’s an example configuration:

      server {
          server_name .foo.net;
      
          include local/tls-common;
      
          charset utf-8;
      
          access_log /var/log/nginx/foo.net-https-access.log;
           error_log /var/log/nginx/foo.net-https-error.log;
      
          # Bypass Fossil for the static Doxygen docs
          location /doc/html {
              root /var/www/foo.net;
      
              location ~* \.(html|ico|css|js|gif|jpg|png)$ {
                  expires 7d;
                  add_header Vary Accept-Encoding;
                  access_log off;
              }
          }
      
          # Redirect everything else to the Fossil instance
          location / {
              include scgi_params;
              scgi_pass 127.0.0.1:12345;
              scgi_param HTTPS "on";
              scgi_param SCRIPT_NAME "";
          }
      }
      server {
          server_name .foo.net;
          root /var/www/foo.net;
          include local/http-certbot-only;
          access_log /var/log/nginx/foo.net-http-access.log;
           error_log /var/log/nginx/foo.net-http-error.log;
      }

Notice that we need two `server { }` blocks: one for HTTPS service, and
one for HTTP-only service:


### HTTP over TLS (HTTPS) Service

The first `server { }` block includes this file, `local/tls-common`:

      listen 443 ssl;

      ssl_certificate     /etc/letsencrypt/live/example.com/fullchain.pem;
      ssl_certificate_key /etc/letsencrypt/live/example.com/privkey.pem;

      ssl_dhparam /etc/letsencrypt/ssl-dhparams.pem;

      ssl_protocols TLSv1 TLSv1.1 TLSv1.2;
      ssl_ciphers "ECDHE-ECDSA-CHACHA20-POLY1305:ECDHE-RSA-CHACHA20-POLY1305:ECDHE-ECDSA-AES128-GCM-SH
      ssl_session_cache shared:le_nginx_SSL:1m;
      ssl_prefer_server_ciphers on;
      ssl_session_timeout 1440m;

These are the common TLS configuration parameters used by all domains
hosted by this server. Since all of those domains share a single TLS
certificate, we reference the same `example.com/*.pem` files written out
by Certbot here. We also reference the common server-specific
Diffie-Hellman parameter file written by the Let’s Encrypt package when
it’s initially installed.

The `ssl_protocols` and `ssl_ciphers` lines are prone to bit-rot: as new
attacks on TLS and its associated technologies are discovered, this
configuration is likely to need to change. Even if we fully succeed in
[keeping this document up-to-date](#evolution), the nature of this guide
is to recommend static configurations for your server. You will have to
keep an eye on this sort of thing and evolve your local configuration as
the world changes around it.

Running a TLS certificate checker against your site occasionally is a
good idea. The most thorough service I’m aware of is the [Qualys SSL
Labs Test][qslt], which gives the site I’m basing this guide on an “A”
rating at the time of this writing.

I assume you’re taking my advice to serve only the least amount of stuff
over HTTP that you can get away with. Certbot’s ACME HTTP-01
authentication scheme is one of those few things.


### HTTP-Only Service

While we’d prefer not to offer HTTP service at all, we need to do so for
two reasons, one temporary and the other going forward indefinitely.

First, until we get Let’s Encrypt certificates minted and configured
properly, we can’t use HTTPS yet at all.

Second, the Certbot ACME HTTP-01 challenge used by the Let’s Encrypt
service only runs over HTTP, because it has to work before HTTPS is
working, or after a certificate is accidentally allowed to lapse.  This
is the protocol Let’s Encrypt uses to determine whether we actually have
control over the domains we want our certificate to be minted for.
Let’s Encrypt will not just let you mint certificates for `google.com`
and `paypal.com`!

So, from the second `service { }` block, we include this file to set up
the minimal HTTP service we reqiure, `local/http-certbot-only`:

      listen 80;
      listen [::]:80;
  
      # This is expressed as a rewrite rule instead of an "if" because
      # http://wiki.nginx.org/IfIsEvil
      #rewrite ^(/.well-known/acme-challenge/.*) $1 break;
  
      # Force everything else to HTTPS with a permanent redirect.
      #return 301 https://$host$request_uri;

As written above, this configuration does nothing other than to tell
nginx that it’s allowed to serve content via HTTP on port 80 as well.

We’ll uncomment the `rewrite` and `return` directives below, when we’re
ready to begin testing.


#### Why the Repitition?

You need to do much the same sort of thing as above for each domain name
hosted by your nginx server.

You might being to wonder, then, why I haven’t factored some of those
directives into the included files `local/tls-common` and
`local/http-certbot-only`. For example, why can’t the second HTTP-only
`server { }` block above just be these two lines:

      server_name .foo.net;
      include local/http-certbot-only;

Then in `local/http-certbot-only`, we’d like to say:

      root /var/www/$host;
      access_log /var/log/nginx/$host-http-access.log;
       error_log /var/log/nginx/$host-http-error.log;

Sadly, nginx doesn’t allow variable subtitution into any of these
directives. As I understand it, allowing that would make nginx slower,
so we must largely repeat these directives in each HTTP `server { }`
block.

These configurations are, as shown, as small as I know how to get them.
If you know of a way to reduce some of this repitition, [I solicit your
advice][fd].


## Step 3: Dry Run

We want to first request a dry run, because Let’s Encrypt puts some
rather low limits on how often you’re allowed to request an actual
certificate.  You want to be sure everything’s working before you do
that.  You’ll run a command something like this:

      $ sudo certbot certonly --webroot --dry-run \
         --webroot-path /var/www/example.com \
             -d example.com -d www.example.com \
             -d example.net -d www.example.net \
         --webroot-path /var/www/foo.net \
             -d foo.net -d www.foo.net

There are two key options here.

First, we’re telling Certbot to use its `--webroot` plugin instead of
the automated `--nginx` plugin. With this plugin, Certbot writes the
ACME HTTP-01 challenge files to the static web document root directory
behind each domain.  For this example, we’ve got two web roots, one of
which holds documents for two different second-level domains
(`example.com` and `example.net`) with `www` at the third level being
optional.  This is a common sort of configuration these days, but you
needn’t feel that you must slavishly imitate it; the other web root is
for an entirely different domain, also with `www` being optional.  Since
all of these domains are served by a single nginx instance, we need to
give all of this in a single command, because we want to mint a single
certificate that authenticates all of these domains.

The second key option is `--dry-run`, which tells Certbot not to do
anything permanent.  We’re just seeing if everything works as expected,
at this point.


### Troubleshooting the Dry Run

If that didn’t work, try creating a manual test:

      $ mkdir -p /var/www/example.com/.well-known/acme-challenge
      $ echo hi > /var/www/example.com/.well-known/acme-challenge/test

Then try to pull that file over HTTP — not HTTPS! — as
`http://example.com/.well-known/acme-challenge/test`. I’ve found that
using Firefox and Safari is better for this sort of thing than Chrome,
because Chrome is more aggressive about automatically forwarding URLs to
HTTPS even if you requested “`http`”.

In extremis, you can do the test manually:

      $ telnet foo.net 80
      GET /.well-known/acme-challenge/test HTTP/1.1
      Host: example.com

      HTTP/1.1 200 OK
      Server: nginx/1.14.0 (Ubuntu)
      Date: Sat, 19 Jan 2019 19:43:58 GMT
      Content-Type: application/octet-stream
      Content-Length: 3
      Last-Modified: Sat, 19 Jan 2019 18:21:54 GMT
      Connection: keep-alive
      ETag: "5c436ac2-4"
      Accept-Ranges: bytes

      hi

You’re looking for that “hi” line at the end and the “200 OK” response
here. If you get a 404 or other error response, you need to look into
your web server logs to find out what’s going wrong.

Note that it’s important to do this test with HTTP/1.1 when debugging a
name-based virtual hosting configuration like this, if the test domain
is one of the secondary names, as in the example above, `foo.net`.

If you’re still running into trouble, the log file written by Certbot
can be helpful.  It tells you where it’s writing it early in each run.



## Step 4: Getting Your First Certificate

Once the dry run is working, you can drop the `--dry-run` option and
re-run the long command above.  (The one with all the `--webroot*`
flags.) This should now succeed, and it will save all of those flag
values to your Let’s Encrypt configuration file, so you don’t need to
keep giving them.



## Step 5: Test It

Edit the `local/http-certbot-only` file and uncomment the `redirect` and
`return` directives, then restart your nginx server and make sure it now
forces everything to HTTPS like it should:

      $ sudo systemctl restart nginx

Test ideas:

*   Visit both Fossil and non-Fossil URLs

*   Log into the repo, log out, and log back in

*   Clone via `http`: ensure that it redirects to `https`, and that
    subsequent `fossil sync` commands go directly to `https` due to the
    301 permanent redirect.

This forced redirect is why we don’t need the Fossil Admin → Access
"Redirect to HTTPS on the Login page" setting to be enabled.  Not only
is it unnecessary with this HTTPS redirect at the front-end proxy level,
it would actually [cause an infinite redirect loop if
enabled](./ssl.wiki#rloop).


## Step 6: Renewing Automatically

Now that the configuration is solid, you can renew the LE cert and
restart nginx with two short commands, which are easily automated:

      sudo certbot certonly --webroot
      sudo systemctl restart nginx

I put those in a script in the `PATH`, then arrange to call that
periodically.  Let’s Encrypt doesn’t let you renew the certificate very
often unless forced, and when forced there’s a maximum renewal counter.
Nevertheless, some people recommend running this daily and just letting
it fail until the server lets you renew.  Others arrange to run it no
more often than it’s known to work without complaint.  Suit yourself.


-----------

<a id=”evolution”></a>
**Document Evolution**

TLS and web proxying are a constantly evolving technology. This article
replaces my [earlier effort][2016], which had whole sections that were
basically obsolete within about a year of posting it. Two years on, and
I was encouraging readers to ignore about half of that HOWTO.  I am now
writing this document about 3 years later because Let’s Encrypt
deprecated key technology that HOWTO depended on, to the point that
following that old HOWTO is more likely to confuse than enlighten.

There is no particularly good reason to expect that this sort of thing
will not continue to happen, so this effort is expected to be a living
document.  If you do not have commit access on the `fossil-scm.org`
repository to update this document as the world changes around it, you
can discuss this document [on the forum][fd].  This document’s author
keeps an eye on the forum and expects to keep this document updated with
ideas that appear in that thread.

[2016]: https://www.mail-archive.com/fossil-users@lists.fossil-scm.org/msg22907.html
[cb]:   https://certbot.eff.org/
[cbnu]: https://certbot.eff.org/lets-encrypt/ubuntubionic-nginx
[fd]:   https://fossil-scm.org/forum/forumpost/XXXXXXXX
[qslt]: https://www.ssllabs.com/ssltest/
[scgi]: https://en.wikipedia.org/wiki/Simple_Common_Gateway_Interface
[vps]:  https://en.wikipedia.org/wiki/Virtual_private_server