SPOPS::Manual::Security - SPOPS security system and how you can customize


This part of the SPOPS manual deals with the last 'S' in SPOPS -- Security. It is one of the main features that sets SPOPS apart from other serialization schemes as well as one of the most confusing. Hopefully we'll be able to clear up any confusion and provide some concrete examples.

This document should answer the following questions:

  • How does security work?

  • How does security use users and groups?

  • How does SPOPS implement security?

  • How can I implement security?

  • How can I customize security?


Security is implemented with a number of methods that are called within the SPOPS implementation module. For instance, every time you call fetch() on an object, SPOPS first determines whether you have rights to do so. Similar callbacks are located in save() and remove(). Unmodified and uninformed of how your users and groups work, SPOPS always allows all actions. You will need to let SPOPS know about your users and groups before you can use security.

We use the Unix-style of permission scheme, separating the scope into: USER, GROUP and WORLD from most- to least-specific. (This is abbreviated as U/G/W.) When we check permissions, we check whether a security level is defined for the most-specific item first, then work our way up to the least_specific. (We use the term 'scope' frequently in the module and documentation -- a 'specific scope' is a particular user or group, or the world.)

Even though we use the U/G/W scheme from Unix, we are not constrained by its history. There is no strict 'ownership' assigned to an object as there is to a Unix file. Instead, an object can have assigned to it permissions from any number of users, and any number of groups.

There are four levels for any object combined with a specific scope:

 NONE:    The scope is barred from even seeing the object.
 SUMMARY: The scope can see an object, but possibly not all of it.
 READ:    The scope can read the object but not save it.
 WRITE:   The scope can read, write and delete the object.

(To be explicit: WRITE permission implies READ permission as well; if a specific scope has WRITE permission for an object, that specific scope can do anything with the object, including remove it.)

Note that the SUMMARY level is not required to be implemented, and many applications have no need of it. We skip it in most discussions below.

Security Rules

With security, there are some important assumptions. These rules are laid out here.

  • The most specific security wins. This means that you might have set permissions on an object to be SEC_LEVEL_WRITE for SEC_LEVEL_WORLD, but if the user who is logged in has SEC_LEVEL_NONE, permission will be denied.

  • All objects must have a WORLD permission. Configuration for your SPOPS object must include the initial_security hash. The only required field is 'WORLD', which defines the default WORLD permission for newly-created objects. If you do not include this, the system will automatically set the WORLD permission to SEC_LEVEL_NONE, which is probably not what you want.

For instance, look at an object that represents a news notice posted:

 Object Class: MyApp::News
 Object ID:    1625

 | USER  | 71827    |        |   X    |         |
 | USER  | 6351     |   X    |        |         |
 | USER  | 9182     |        |        |    X    |
 | GROUP | 762      |        |   X    |         |
 | GROUP | 938      |        |        |    X    |
 | WORLD |          |        |   X    |         |

From this, we can say:

  • User 6351 can never view this notice. Even though the user might be a part of a group that can; even though WORLD has READ permission. Since the user is explicitly forbidden from viewing the notice, nothing else matters.

  • If a different User (say, 21092) who belongs to both Group 762 and Group 938 tries to determine permission for this object, that User will have WRITE permission since the system returns the highest permission granted by all Group memberships.

  • Any user who is not specified here and who does not belong to either Group 762 or Group 938 will get READ permission to the object, using the permission for the scope WORLD.

Setting Security: User and Group Objects

It is a fundamental tenet of this persistence framework that it should have no idea what your application looks like. However, since we deal with user and group objects, it is necessary to enforce some standards.

  • Must be able to retrieve the ID of the object with the method call 'id'. The ID value can be numeric or it can be a string.

  • Must be able to get an arrayref of members. With a group object, you must implement a method that returns users called 'user'. Similarly, your user object must implement a method that returns the groups that user belongs to via the method 'group':

      1:  # Note that 'login_name' is not required as a parameter; this is just
      2:  # an example
      4:  my $user_members = eval { $group->user };
      5:  foreach my $user ( @{ $user_members } ) {
      6:      print "Username is $user->{login_name}\n";
      7:  }
      9:  # Note that 'name' is not required as a parameter; this is just an
     10:  # example
     12:  my $groups = eval { $user->group };
     13:  foreach my $group ( @{ $groups } ) {
     14:      print "Group name is $group->{name}\n";
     15:  }
  • Must be able to retrieve the logged-in user (and, by the rule stated above, the groups that user belongs to). This is done via the global_user_current() method call. The SPOPS object or other class must be able to fulfill this method and return a user object.


An object moving from the non-serialized to the saved state is a special case for security. We cannot determine in our usual manner what security the current user has because the object has not yet been created. Generally, we rely on the application to determine whether the user should be able to create an object at all. Once we get past that hurdle, we just need to figure out what permissions the object should have when it's first created. After that, we're set.

The process for determining what security a newly created object should have can be simple, or it can be complicated :) It is designed to be flexible enough for us to easily plug-in security policy modules whenever we write them, but simple enough to be used just from the object configuration.

Object security configuration information is specified in the 'creation_security' hashref in the object configuration. A typical setup might look like:

  creation_security => {
     u   => undef,
     g   => { 3 => 'WRITE' },
     w   => 'READ',

Each of the keys maps to a (hopefully intuitive) scope:


For each scope you can either name security specifically or you can defer the decision-making process to a subroutine. The former is called 'exact specification' and the latter 'code specification'. Both are described below.

Note that the 'level' values used ('WRITE' or 'READ' above) do not match up to the SEC_LEVEL_* values exported from this module. Instead they are just handy mnemonics to use -- just lop off the 'SEC_LEVEL_' from the exported variable:


Exact specification

'Exact specification' does exactly that -- you specify the ID and security level of the users and/or groups, along with one for the 'world' scope if you like. This is handy for smaller sites where you might have a small number of groups.

The exact format is:

 User and World



   SCOPE => ID => LEVEL,
            ID => LEVEL, ... } }

Where 'SCOPE' is 'u' or 'g', 'ID' is the ID of the group/user and 'LEVEL' is the level you want to assign to that group/user. So using our example above:

     g   => { 3 => 'WRITE' },

We assign the security level SEC_LEVEL_WRITE to the group with ID 3.

For the SEC_SCOPE_USER scope, if you specify a level:

    u    => 'READ',

Then that security level is assigned for the user who created the object.

If you specify anything other than a level for the SEC_SCOPE_WORLD scope, the system will discard the entry and assign it the SEC_LEVEL_NONE level.

Code specificiation

You can also assign the entire process off to a separate routine:

  creation_security => {
     code => [ 'My::Package' => 'security_set' ]

This code should return a hashref formatted like this

   u => SEC_LEVEL_*,
   g => { gid => SEC_LEVEL_* },
   w => SEC_LEVEL_*

If you do not include a scope in the hashref, no security information for that scope will be entered. (Except for the world scope, which will get a SEC_LEVEL_NONE if it's not specified.)


SPOPS comes with one implementation for security objects, SPOPS::Security::DBI. Implementations of the security object must implement the following methods.

fetch_by_object( $object, \%params )

Find all security levels for a particular object and scope.

You can restrict the security returned for USER and/or GROUP by passing an arrayref of objects or ID values under the 'user' or 'group' keys.



A user object or ID.


An arrayref of group of objects or group IDs.


If you do not pass in an $object, you can specify it by its class and ID. This should be a full object class name.


If you do not pass in an $object, you can specify it by its class and ID. This should be a full object ID.


 my \%info = $sec->fetch_by_object( $obj );

Returns all security information for $obj.

 my \%info = $sec->fetch_by_object( $obj, { user  => 2,
                                            group => [ 817, 901, 716 ] } );

Returns $obj security information for WORLD, USER 2 and GROUPs 817, 901, 716.

 my $current_user = My::Object->global_user_current;
 my \%info = $sec->fetch_by_object( undef, { class     => 'My::Object',
                                             object_id => 'dandelion',
                                             user      => $user,
                                             group     => $user->group } );

Returns security information for the object of class My::Object with the ID dandelion for the current user and the user's groups.

Returns: a hashref with security information for $object for a given scope. The keys of the hashref are SEC_SCOPE_WORLD, SEC_SCOPE_USER, and SEC_SCOPE_GROUP as exported by SPOPS::Secure.

fetch_match( \%params )

Returns a security object matching the $obj for the scope and scope_id passed in, undef if none found.


 my $sec_class = 'My::Security';

 # Returns security object matching $obj with a scope of WORLD

 my $secw = $sec_class->fetch_match( $obj,
                                     { scope => SEC_SCOPE_WORLD } );

 # Returns security object matching $obj with a scope of GROUP
 # matching the ID from $group
 my $secg = $sec_class->fetch_match( $obj,
                                     { scope    => SEC_SCOPE_GROUP,
                                       scope_id => $group->id } );

 # Returns security object matching $obj with a scope of USER
 # matching the ID from $user
 my $secg = $sec_class->fetch_match( $obj,
                                     { scope    => SEC_SCOPE_USER,
                                       scope_id => $user->id );


The SPOPS security scheme is flexbile enough for you to implement your own security. For instance, if you had a database of contacts for your national membership organization you might want to ensure that each state sees only the contacts within its state.

To do this, you could simply create a get_security() method in your contact class. A simplified example of what such a method might look something like:

 sub get_security {
     my ( $self, $p ) = @_;
     my $log = get_logger();
     my ( $user, $group_list ) = $self->get_security_scopes( $p );
     if ( my $security_info = $self->_check_superuser( $user, $group_list ) ) {
         $log->is_info &&
             $log->info( "Superuser is logged in, can do anything" );
         return $security_info;
     if ( $self->{state} eq $user->{state} ) {
         return { SEC_SCOPE_WORLD() => SEC_LEVEL_WRITE };
     return { SEC_SCOPE_WORLD() => SEC_LEVEL_NONE };

For a good example of what you can do with subclassing, see the code for the subclass SPOPS::Secure::Hierarchy.


Copyright (c) 2001-2004 Chris Winters. All rights reserved.

See SPOPS::Manual for license.


Chris Winters <>