1. Creating a Certificate Authority.

  Before running ca-init(1), a configuration file for the CA scripts must be
created. This configuration file sets up some templating variables that will
be present in certificates created for this CA, such as the domain, CA name,
and the root directory which will be populated with the generated certificates.
An example configuration file is provided with the scripts, and the comments
should be self-explanatory. 

  By default the CA scripts will read /etc/ca-scripts.conf. This is fine for
creating a single CA serving a single domain with no intermediary certificates,
but for a more complex setup a directory of configuration files will probably
be needed. Some settings are required, namely the CA_HOME, CA_DOMAIN, and
CA_DN_* variables, while others can be inferred from these or have sensible
defaults set. See ca-scripts.conf(5) for more detail on these.

  Once the configuration has been created the initial CA setup can be performed
with ca-init(1), but please note that the path set in CA_HOME must exist and be
writeable before it will run correctly. It is recommended (but not in any way
required) to create an unprivileged "ssl" user to run all the scripts as, so
the permissions are correctly set. A number of subdirectories will be set
up underneath this root, and an openssl configuration file, certificate and
private key will be generated. This key can be 3DES encrypted for security.

  Optionally, it is possible to split the initial setup process so that the
directory structure and openssl configuration generation can be done in a
seperate step to the generation of the CA certificates, so that the config can
be manually edited. To fully understand it's contents you're unfortunately
going to need to read ca(1ssl), req(1ssl), x509(1ssl), config(5ssl), and
x509v3_config(5ssl). Particularly important are the x509v3 extensions present
in the certificate, which are defined in the "ca_x509_extensions" section of
the config file.

2. Creating a certificate.

  The ca-create-cert(1) script can generate three "types" of certificate: 
server certificates for securing a service with SSL/TLS; client certificates
for authenticating a client to these services; and user certificates for
authentication, S/MIME e-mail signing or encryption, and code signing. There
are minor but important differences in the key usage extensions present in
these different certificate types, details can be found in the extension
templates provided with the scripts.

  ca-create-cert(1) takes a number of options to customise the generated 
certificate. The --type option is mandatory, and for server certs it is very
likely that the --alt-name option will be useful to set x509v3 SubjectAltName
DNS records for other hostnames for the server. Both the server hostname and
any alternative names will be fully-qualified to CA_DOMAIN if they do not
contain any dots, but if unqualified names are passed in they are also
preserved as alternative DNS names in the certificate. The private key may be
encrypted with 3DES, and optionally the certificate, key, and CA certificate
can be bundled together into a PKCS#12 format certificate archive. By default
certificates are valid for 365 days from signing, but this may be changed with
the --days option.

  The certificate's DN can be completely changed from the defaults provided by
ca-scripts.conf(5), but be wary as by default the generated openssl config file
requires that the country (C) and organisation (O) fields match those of the CA
certificate. A comment may also be set that will show up in user browsers when
they click on their padlock icons to examine the certificate's properties. As 
with the CA setup, the steps to generate the certificate can be split up so
that configurations that are created from templates can be edited beforehand.

3. Renewing a certificate.

  Certificates are renewed using ca-renew-cert(1). This script currently does
some painful certificate manipulation that is not strictly necessary in most
cases, and may in fact decrease SSL security slightly. This is done because
the normal renewal process re-generates the certificate signing request and
thus creates a new public/private keypair. If the certificates are used for
S/MIME encryption or code signing, this renders all the encrypted e-mail
unreadable and requires you to re-sign the code with your new private key.

  To avoid this, ca-renew-cert(1) re-signs the old certificate request with a
a new expiry date using the extensions generated when the old certificate was
signed. In the future it is possible (even likely) that this renewal method
will only be used on "user" type certificates, and the "server" and "client"
types will be renewed normally. If the current renewal method doesn't provide
sufficient security, the current certificate should be revoked and a new one
generated that is valid for the correct period of time using the --days option
to ca-create-cert(1).

  As with the certificate creation script the --type option is mandatory for
ca-renew-cert(1), but the argument may be either a hostname, a username or a
path to a certificate. Internally this will be resolved to the correct
information required for certificate renewal.

4. Revoking a certificate.

  Revoking a certificate is done by giving the hostname, username or path to
the certificat to revoke-cert.sh. This script also regenerates a new CRL in
both PEM and DER encodings (firefox prefers the latter while IE and other
browsers work better with the former), and re-generates the html file with the
new fingerprints.