Sophie

Sophie

distrib > Mageia > 7 > armv7hl > media > core-release > by-pkgid > 9f7c3bc318f0e910c57eb4772665f20d > files > 13

pwdb-conf-0.62-18.mga7.armv7hl.rpm

  The PWDB Library Guide
  Cristian Gafton and Andrew G. Morgan
  DRAFT v0.54 1996/12/4

  This manual documents the Password database library (libpwdb). This is
  a library implementing a generic user information database. It was
  specifically designed to satisfy the needs of Linux-PAM.
  ______________________________________________________________________

  Table of Contents


  1. Introduction
     1.1 Motivation

  2. Configuration
     2.1 The configuration file

  3. Implementation
     3.1 The core (public) functions
        3.1.1 Initialization and termination
        3.1.2 Structure management
        3.1.3 Database query functions
        3.1.4 Diagnostics
        3.1.5 Types of database
     3.2 Functions supplied by database modules
     3.3 Standard entries in pwdb structures
     3.4 Helper functions
        3.4.1 Public functions
        3.4.2 Private functions
     3.5 Typical usage

  4. Supported modules
     4.1 UNIX module
     4.2 Shadow module
     4.3 NIS module
     4.4 DECNIS module
     4.5 RADIUS module

  5. Proposals/unresolved issues
  6. References
  7. Acknowledgments


  ______________________________________________________________________

  1.  Introduction

  This is a document describing the libpwdb functions (or how they will
  function).  Cristian and Andrew will be modifying this file as we go
  along. Hopefully it will describe the library when it is finished.


  1.1.  Motivation

  Linux, at the level of the kernel, identifies a user by a 16-bit
  integer--their UID. At the level of a functioning system, the user has
  other characteristics; a name, group memberships, a home, a number of
  aliases (email address etc.) and a password.


  The mapping of a UID to each of these pieces of personal data is with
  respect to a database file.  Unfortunately, depending on circumstances
  there is no single location for this database.  The most basic Linux
  system might have a single /etc/passwd file, with a single entry for
  each user, and an /etc/group file containing a list of the groups
  defined on the local system.  In shadow aware systems a user might
  additionally have an /etc/shadow entry.  In a networked environment,
  all of the users might be identified by entries in a file located on a
  remote computer.


  While it is possible for a local system to adopt one scheme to satisfy
  the demands of the POSIX getpw.. commands, it is problematic for an
  individual system to simultaneously support a variety of databases to
  define its users and groups.


  It is desirable for applications to be able to obtain user-information
  in a transparent manner.  At the same time certain applications might
  like to obtain information from a specific database for a given user -
  these applications might perform the task of maintaining and updating
  databases for example.  The functions of libc, do not currently offer
  this flexibility.  It is for this reason that libpwdb has been
  created.  It is intended to provide a flexible and yet simple database
  management primarily for user and group information under Linux
  systems.  An API that all applications can use.


  Should this library become popular, it is anticipated that the
  database will be extended to include other types of database files.


  2.  Configuration

  2.1.  The configuration file

  Use of libpwdb requires the presence of a correctly formatted
  configuration file: /etc/pwdb.conf. It has the following syntax:



           #
           # first the list of user databases
           #
               user:
                       list1
                       list2
                       list3
                       ...
           #
           # next, the list of group databases
           #
               group:
                       list1
                       list2
                       ...
           #
           # end of file
           #



  Here, listN has the following form name1+name2+name3. It indicates a
  collection of databases that are merged to form the record for the
  user/group.  When selecting the default database (PWDB_DEFAULT) the
  library chooses the list whose first named database contains an entry
  for the requested user (id). Currently, the nameN items are from the
  following selection: nis, unix, radius and shadow.
  Newlines are ignored except where they terminate comments; comments
  are preceded with `#' characters.


  A simple example /etc/pwdb.conf file would be:


           # This is an example /etc/pwdb.conf file. It defines the
           # database information sources for users in this system

           # First, we define where user information is stored
           # (here, users are listed in /etc/passwd and may have
           # supplementary information in the /etc/shadow file)
           # Should the user have a shadow entry the combination is
           # used

               user:
                       shadow + unix
                       unix

           # Second, we define where the users' groups are listed
           # (here the groups are listed in the /etc/group file
           # in addition, there is group information to be found
           # in the /etc/sgroup file)

               group:
                       unix + shadow

           #
           # end of file



  3.  Implementation

  This will get completed as we develop the library


  3.1.  The core (public) functions

  libpwdb offers the following generic interface:


  3.1.1.  Initialization and termination


  +o  int pwdb_start(void)


     Initialize the library for use by the current application Will read
     the configuration file and publicize the database policies (as they
     are listed in /etc/pwdb.conf) in const int *pwdb_policy and const
     int *pwdb_group_policy.  The library maintains a count of the
     number of times it has been pwdb_start()ed.


  +o  int pwdb_end(void)


     Once the pwdb_start()ed count returns to zero, this function closes
     down the library and free()'s all memory allocated by it.  Any
     attempt to pwdb_end() the library more times than it has been
     pwdb_start()ed will cause PWDB_ABORT to be returned. This feature
     can be used by an application to guarantee the library is
     shutdown..


           while (pwdb_end() == PWDB_SUCCESS);



  3.1.2.  Structure management


  +o  int pwdb_entry_delete(const struct pwdb_entry **e)


     free() the memory associated with the pointer *e. Its value will be
     overwritten with '\\0's before the memory is free()d. This is
     reassuring from the point of view of minimizing security problems.
     This function should be used to liberate a pwdb_entry returned by
     pwdb_get_entry().


  +o  int pwdb_get_entry(const struct pwdb *p, const char *entry, const
     struct pwdb_entry **e);


     Read (and duplicate) an entry from the argument pwdb structure.
     Should the requested entry not prove to be present, PWDB_NOT_FOUND
     is returned.


  +o  int pwdb_set_entry(const struct pwdb *p, const char *entry, const
     void *datum, const int length, int (*compare)(const void *, const
     void *, int), int (*strvalue)(const void *, char *, int), int
     max_strval_size))


     Set an entry in the argument pwdb structure. One can delete an
     entry from a struct pwdb by calling this function with NULL for
     datum and -1 for length.


     The two functions passed as arguments in this call are as follows:



     int compare(const void *value1, const void *value2, int length)
        compare the two entry values. They are both of given length in
        chars.


     int strvalue(const void *value, char *buffer, int length)
        Produce a text version of the given entry value. The buffer is
        the destination of the output text and the length is that of the
        *value in bytes. Note, the buffer is guaranteed by the calling
        process to be long enough for the output text.



  +o  int pwdb_delete(const struct pwdb **old)


     Applications use this to liberate the memory associated with a pwdb
     structure.  Following a successful free(), *old is set to NULL.

  +o  int pwdb_new(const struct pwdb **new, int life_sec)


     Applications can request a new (blank) pwdb structure with this
     call. Note that it is returned as a const structure.  This is to
     prevent the application from altering the structure directly. An
     application should use the library functions to modify this
     structure.  The life_sec time is used to limit the length of time
     this pwdb structure will be valid for. It is some number of seconds
     from the present. If life_sec is non-zero, the pwdb structure will
     expire in that many seconds. Zero indicates the pwdb structure will
     non expire.



       int pwdb_merge(const struct pwdb *target, const struct pwdb *source,
                      int overwrite)



  this function copies the elements of source to target. If overwrite is
  PWDB_TRUE then all elements of source that are also in target will be
  overwritten by this call.



  +o  int pwdb_expire(const struct pwdb *p, int life_sec)

     This function can shorten the lifetime of a the referenced struct
     pwdb. It computes the expiry time from the present with respect to
     life_sec and if this is before the expiration time currently
     associated with the *p structure, it shortens the life of the
     structure accordingly.


     Note, it is not possible to extend the life of a pwdb structure,
     only to shorten it. An argument of 0 or less will result in the
     immediate expiry of the pwdb structure.


  +o  int pwdb_source(const struct pwdb *old, const pwdb_type *src, const
     char *class, const char *name, const int id)

     This function is used to set the source of the indicate pwdb
     structure. The argument src is a pointer to a list of pwdb_type
     entries. This list is terminated with a _PWDB_MAX_TYPES item.
     Valid types are listed in the <security/pwdb_public.h> file. The
     remaining arguments are used to initialize the caching facilities.


  3.1.3.  Database query functions


  +o  int pwdb_locate(const char *class, const pwdb_type *src, const char
     *name, const int id, struct pwdb **p)


     Obtain the entry for a given name and/or id in the indicated
     database. If *p is not NULL the database-module may choose to use
     the information it contains. It is intended that information
     obtained with this function is merged with that of the original *p
     structure. If *p is NULL, then a struct pwdb is allocated by the
     pwdb_locate() function in the first module used (and eventually
     merged with the subsequent modules, depending on the local setup).
     The class is the kind of database we are searching, examples
     include user and group.


  +o  int pwdb_request(const char *class, const pwdb_type *src, const
     char *entry, struct pwdb **p)


     This function will further query the database, for an entry that
     may depend on the entries already present in the *p structure. The
     entrys that can be appended to the existing *p structure are
     defined for the class of database.  For example, one may request
     the "groupids" entry from "group" class, which will search the
     group database for the list of groups to which a given user
     belongs. The name of the user is passed as an entry in the
     preexisting pwdb structure.


  +o  int pwdb_replace(const char *class, const pwdb_type *src, const
     char *name, const int id, struct pwdb **p)


     Add/replace the entry for a name and/or id in the indicated list of
     databases. The fields for the new database entries are taken from
     the p argument. The src argument is the list of entries that will
     be updated.


  +o  int pwdb_remove(const char *class, const pwdb_type *src, const char
     *name, const int id, struct pwdb **p)


     Remove the entry for the indicated name and/or id in the indicated
     database. If not NULL the remove function may obtain access
     information from the p argument.


     Note, this function only acts on the specified class of database.
     If reference to a name or id is present in another class of
     database, then it is the responsibility of the application to purge
     these databases too. The pwdb_remove() function is not sufficiently
     powerful to follow up on cross-references.


     As an example of the above concern, consider the removal of name
     ``joe'' from the "user" database. The "group" database entries that
     refer to ``joe'' as a group-member are unaffected by this
     pwdb_remove() request.  Instead, it is the responsibility of the
     calling application to search for such entries and systematically
     pwdb_remove() them.


  +o  int pwdb_support(const char *class, const pwdb_type *src, const
     char *entry_name)


     Indicate whether the given entry name is supported by the given
     database(s). PWDB_SUCCESS indicates yes. Anything else is NO (or
     something more specific, could be ``yes, if you supply a
     pass_phrase'' for example).


  +o  int pwdb_flags(const char *class, const pwdb_type *db, pwdb_flag
     *flag_p)

     In order to know in advance whether a process is able to
     read/modify a specified database, this command is provided by each
     module.  The input arguments are the class of database (user, group
     etc.), db (the database(s) we are going to use) and some storage
     space for the returned flags.


     Valid return flags which can be logically OR'd together are:



     PWDB_F_NOACCESS
        insufficient privilege to access database


     PWDB_F_NOUPDATE
        insufficient privilege to alter an entry


     PWDB_F_PASS_PHRASE
        to access the database, the process must supply a "pass_phrase"
        entry with a preallocated pwdb structure (use pwdb_new() call)


     ..and..
        more flags are likely to be added



     To establish if a given flag is set the following macro is
     supplied:


  +o  pwdb_on(flag, PWDB_F_XXX)


     it returns TRUE(1) or FALSE(0).



  3.1.4.  Diagnostics


  +o  const char *pwdb_strerror(int pwdb_error)


     return a textual description of the indicated return value.


     Valid return codes are:



     PWDB_SUCCESS
        task completed successfully


     PWDB_BAD_REQUEST
        request not recognized


     PWDB_TOO_WEAK
        insufficient privilege for operation


     PWDB_ABORT
        internal failure - seek help


     PWDB_BLOCKED
        another process has locked resource


     PWDB_MALLOC
        insufficient memory for operation


     PWDB_NOT_FOUND
        requested item was not found


     PWDB_PASS_PHRASE_REQD
        pass_phrase needed to satisfy request here the application
        should supply a pwdb structure with a "pass_phrase" entry and
        call the function again


     PWDB_CONF_ERR
        there is a problem with the PWDB_CONF file.


     PWDB_EXPIRED
        the referenced pwdb structure has expired it is no longer valid
        and should be deleted.


     PWDB_UNSUPPORTED
        this function is not supported by some module (not supported
        means also unimplemented, for a while...)


     PWDB_TIMEOUT
        a timeout occured while performing the function. Presently could
        show up only when using RADIUS interface.



  3.1.5.  Types of database

  The pwdb_type of database a request is associated with is given by one
  of the following values:



     PWDB_DEFAULT
        no database indicated, use configured list


     PWDB_UNIX
        generic /etc/passwd and /etc/group files


     PWDB_SHADOW
        /etc/shadow and /etc/gshadow Intended to supplement other
        databases


     PWDB_NIS
        Use NIS server

     PWDB_RADIUS
        Use RADIUS server


     PWDB_DECNIS
        Use a NIS server configured for Digital Equipment Corp Enhanced
        Security, or a Solaris server with an adjunct NIS password file.



  +o  const char *pwdb_db_name(pwdb_type src)


     return a character representation of the database functions



  3.2.  Functions supplied by database modules

  Each module must provide 7 functions to the generic pwdb interface.
  They are registered with the generic interface via a structure of the
  following form:



       struct _pwdb_module {
           pwdb_type type;                    /* type of database (code) */
           const char *name;                  /* type of database (text) */
           const char *class;                 /* class of database (user/group) */

           /* FUNCTIONS: used to access the relevant database */

           int (*locate)(const char *name, const int id, const struct pwdb **p);
           int (*request)(const char *entry_name, const struct pwdb **p);
           int (*replace)(const char *name, const int id, const struct pwdb **p);
           int (*delete)(const char *name, const int id, const struct pwdb **p);
           int (*support)(const char *entry_name);
           int (*flags)(pwdb_flag *flags);
           int (*cleanup)(int code);
       };



  For the functions above taking a name and an id entry, the application
  may choose to leave one unspecified with the following defaults:


     PWDB_NAME_UNKNOWN
        just look at the id field


     PWDB_ID_UNKNOWN
        just look at the name field


  In the case that the application supplies neither the name or the id,
  the module functions will try to obtain the relevant information from
  the argument pwdb structure.


  It is legal for both the name and id to be specified. In this case
  they must both match an entry in the database to satisfy one of the
  above function calls. If both values are supplied and there is no
  entry in the database which matches them, PWDB_BAD_REQUEST is
  returned.


  The structure is registered via an entry in the modules list (see
  pwdb_module.c).


  3.3.  Standard entries in pwdb structures

  The following are standard entries in the pwdb structure. They can be
  read/written with calls to pwdb_g/set_entry.


  First, we consider the "user" class of databases. For these, two
  entries are mandatory. They correspond to the name of the user and the
  user's uid.



     user
        character string; the user's login id.


     uid
        uid_t; the user's user-id.



  The next entries are named by convention. Where possible new database
  functions should map these entries into their corresponding fields.
  These entries correspond to the entries in the /etc/passwd file.



     passwd
        character string; the encrypted password for the user.


     defer_pass
        This entry is intended to take care of situations that the
        normal passwd field is not used for the password. The defer_pass
        entry contains a character string that has typically two
        functions:

        +o  it indicates that the password is to be stored in the
           database that sets this entry.

        +o  it serves as a substitution string for databases that would
           normally contain the password entry.


        For example, for a unix+shadow setup, defer_pass would have the
        value ``x''. The unix (no shadow) value for this entry is ``U''
        which implies that the passwd field came from the user's entry
        in the /etc/passwd file.


     gid
        gid_t; the user's principal group-id.


     group
        character string; naming the user's principal group.
     gecos
        character string; giving a more complete name for the user. It
        is conventional for this field to contain office and other
        information concerning the real-world identity of the user.


     dir
        character string; the home directory of the user.


     shell
        character string; the shell that the user prefers to use.



  These entries correspond to the entries in the /etc/group file in
  addition to the user and gid entries above. They can be
  pwdb_request()d from the "group" class of databases.



     groups
        character string; listing the group memberships of the user. The
        field separators are commas -- no spaces.


     groupids
        array of gid_t; an array containing the group id's of the user
        in numerical form.



  The following are intended to correspond to /etc/shadow entries.



     last_change
        long integer; day of last change of password


     min_change
        long integer; minimum number of days between password changes


     max_change
        integer; maximum number of days between password changes


     warn_change
        long integer; number of days to warn user to change their
        password before it expires


     defer_change
        long integer; number of days after a user's password has expired
        before the user is denied access


     expire
        long integer; day the user's account expires


     shadow_flags
        long integer; reserved for use by underlying shadow
        implementation
  The following is the entry used to supply a clear-text password for
  access to the database.



     pass_phrase
        character string; this is the password required to access the
        user's record in a database



  When integrating another database format the implementor is strongly
  encouraged to try to reuse the entries above to the extent they are
  appropriate. Should there be an absent entry in any database, the
  database management functions should be able to supply a reasonable
  default but only when updating its database.


  3.4.  Helper functions

  3.4.1.  Public functions


  +o  char *_pwdb_delete_string(char *s)


     overwrite the string 's' and return NULL. usage:


     old_ptr = _pwdb_delete(old_ptr);


  +o  char *_pwdb_dup_string(const char *s)


     malloc() a copy of the string 's'. Return its address or NULL if s
     == NULL or on error.


     this memory will not be free()'d by a call to pwdb_end().


  +o  void pwdb_print_pwdb_struct(const struct pwdb *p)


     Dump the contents of *p to the stderr. Useful for debugging.



  3.4.2.  Private functions


  +o  static struct pwdb *_pwd_check(const struct pwdb *p)


     Establish if the pwdb structure was allocated by the library This
     library should not honor requests from elsewhere.  return the local
     version (non-const) of this structure or NULL on error.


     this function is strictly designed for the use of the generic code.
     Both modules and applications should never need to call it.



  3.5.  Typical usage

  Here is a skeleton usage for a login type program.



           pwdb_start();
           pwdb_locate("user", PWDB_DEFAULT, username, PWDB_ID_UNKNOWN, &pw);
           pwdb_request_group("group", PWDB_DEFAULT, "groupids", &pw);
           pwdb_get_entry(pw, "uid", &e1);
           pwdb_get_entry(pw, "gid", &e2);
           pwdb_get_entry(pw, "groupids", &e3);
           pwdb_end();



  4.  Supported modules

  PLEASE NOTE. Currently few group functions have been implemented.


  4.1.  UNIX module

  This section documents the current state of the UNIX module. From the
  point of view of the application, it is named "unix" and has the
  pwdb_type PWDB_UNIX.


  Entries supported by the "user" class of the UNIX module are as
  follows:

     user
        - username

     uid
        - user-id

     gid
        - group-id

     passwd
        - encrypted password

     defer_pass
        - "U" unless set by other database

     gecos
        - user information

     dir
        - home directory

     shell
        - shell executable


  Entries supported by the "group" class of the UNIX module are as
  follows:

     group
        - username


     gid
        - group-id

     passwd
        - encrypted password

     users
        - text list of user names separated by commas


  The pwdb_request() function call is only supported for the "group"
  class. The two entries that may be requested are:

     groups
        - text list of group names separated by commas

     groupids
        - array of gid_t values that contain the numerical form of the
        "groups" entry.

  Note, for such requests to be honored the name of the user should be
  contained in the pwdb ** argument prior to the pwdb_request() call.


  4.2.  Shadow module

  This section documents the current implementation of the shadow
  database module.


  The shadow module does not make any use of the id parameter. Since the
  shadow database does not contain such an entry, the only way of
  identifying a user is with a name argument.  However, an id based
  lookup is possible, if the shadow database is searched after a
  database that contains a user-uid mapping. In this case the ``user''
  entry in the partially built pwdb structure is used to locate the
  appropriate entry in the shadow file.


  The shadow module is designed to work in conjunction with a database
  that provides the standard user-uid mapping.  It should be noted that
  it does not provide sufficient information to support a user login
  session.


  Entries supported by the "user" class of the shadow module are:

     user
        - username

     passwd
        - encrypted password

     last_change
        - date password was last changed

     min_change
        - minimum period before password can be changed

     max_change
        - lifetime of current password

     warn_change
        - number of days prior to expiry that the user should be warned


     defer_change
        - grace period before password is finally invalid

     expire
        - date account expires


  Entries supported by the "group" class of the shadow module are:

     group
        - groupname

     passwd
        - encrypted group password

     users
        - text list of user names separated by commas (members of the
        group)

     admins
        - text list of user names separated by commas (administrators of
        the group)



  4.3.  NIS module


  +o  user removal is not possible. (lack of documentation)

  +o  user creation is not possible (lack of documentation)


  4.4.  DECNIS module

  The decnis module allows use of some extensions to the basic NIS maps.
  The two current extensions supported are the Solaris passwd.adjunct
  and the Digital Equipment Corp prpasswd maps. Both of these are shadow
  password schemes for basic (non-secure) NIS. The solaris support is
  untested.  This option encompasses the NIS option, and the decnis
  option should replace the standard nis option in the pwdb.conf file.
  In the event of a user not having a shadow NIS password entry, the
  result should be identical to that returned by the NIS module above.

  Warning1 Support for many of the extensions (password expiry, life
  time, account locking etc..) is not present in this version. Just raw
  authentication.

  Warning2 Clustering your Linux boxes with this software will reduce
  the security level of your NIS server/cluster. Users may have access
  to encrypted passwords of other users.



  +o  user removal is not possible. (lack of documentation)

  +o  user creation is not possible (lack of documentation)


  4.5.  RADIUS module

  The RADIUS module is acting just as a user validation mechanism. The
  official Livingston radiusd 2.0 is supported, but in order to take
  advantage of the all information and auth tokens the radius server can
  provide, a session PAM module should be written, and be stacked over
  pam_unix.


  The following should be taken into consideration when writing
  applications to authenticate to a radius server:



  +o  the pwdb group functions are not supported (RADIUS does not have
     this concept)

  +o  Other than checking for username/password pairs, the radius module
     can not be used alone with the stock radiusd server to handle the
     user login without a suitable RADIUS client (for example there is
     no way to get an UID for a user).

     However, with few hacks to the radius server and proper
     modification of the /etc/raddb/dictionary file this module uses, a
     NIS-like environment could be achieved. Full documentation on how
     to achieve this will be added later.


  +o  All radius module functions that access the remote server require
     the presence of the "pass_phrase" pwdb entry which contains the
     user password in clear text. The radius module will destroy this
     information as soon as it can dispose of it, so an application can
     assume that after a call to a function in the radius module which
     returned with PWDB_SUCCESS, the pass_phrase entry is wiped out.

  +o  When updating a RADIUS user password, one should supply both
     pass_phrase and passwd entries in clear text. By convention, the
     pass_phrase contains the old password (which is required by the
     RADIUS server to authenticate the user) and the passwd entry
     contains the new password.  One should be very careful about this
     issue, as some other modules used before RADIUS authentication may
     set the value of passwd entry, and the application should make sure
     that the clear text password is passed to the RADIUS module for
     changing password. The RADIUS module will wipe out both pass_phrase
     and password entries - thus the application can not rely on
     validity of any of those entries in the pwdb structure after a call
     to the update function of the RADIUS module.


     Note that the RADIUS server must permit changing of the passwords
     user passwords. IF the RADIUS does not accept changing the user
     passwords, a PWDB_TIMEOUT will occur.


  +o  The password update is the only function supported by the RADIUS
     update function.


  Entries supported by this database are set according to the
  definitions from /etc/raddb/dictionary file. Three entries have a
  special meaning when calling the RADIUS functions: user, passwd and
  pass_phrase. The passwd and pass_phrase will be wiped out by the
  RADIUS functions as soon as the module can dispose them. The reponse
  from the RADIUS server is processed and entries are set in the pwdb
  structure according to the names from the dictionary file.


  A sample RADIUS dictionary entry list is provided here - valid for
  Livingston RADIUSD 2.0:



  #---------------------------------------------------------------------------
  #
  # @(#)dictionary        1.3 10/1/96  Copyright 1991 Livingston Enterprises Inc
  #
  #---------------------------------------------------------------------------
  #
  #       This file contains dictionary translations for parsing
  #       requests and generating responses.  All transactions are
  #       composed of Attribute/Value Pairs.  The value of each attribute
  #       is specified as one of 4 data types.  Valid data types are:
  #
  #       string - 0-253 octets
  #       ipaddr - 4 octets in network byte order
  #       integer - 32 bit value in big endian order (high byte first)
  #       date - 32 bit value in big endian order - seconds since
  #                                       00:00:00 GMT,  Jan.  1,  1970
  #

  ATTRIBUTE       User-Name               1       string
  ATTRIBUTE       Password                2       string
  ATTRIBUTE       CHAP-Password           3       string
  ATTRIBUTE       NAS-IP-Address          4       ipaddr
  ATTRIBUTE       NAS-Port                5       integer
  ATTRIBUTE       Service-Type            6       integer
  ATTRIBUTE       Framed-Protocol         7       integer
  ATTRIBUTE       Framed-IP-Address       8       ipaddr
  ATTRIBUTE       Framed-IP-Netmask       9       ipaddr
  ATTRIBUTE       Framed-Routing          10      integer
  ATTRIBUTE       Filter-Id               11      string
  ATTRIBUTE       Framed-MTU              12      integer
  ATTRIBUTE       Framed-Compression      13      integer
  ATTRIBUTE       Login-IP-Host           14      ipaddr
  ATTRIBUTE       Login-Service           15      integer
  ATTRIBUTE       Login-TCP-Port          16      integer
  ATTRIBUTE       Reply-Message           18      string
  ATTRIBUTE       Callback-Number         19      string
  ATTRIBUTE       Callback-Id             20      string
  ATTRIBUTE       Framed-Route            22      string
  ATTRIBUTE       Framed-IPX-Network      23      ipaddr
  ATTRIBUTE       State                   24      string
  ATTRIBUTE       Session-Timeout         27      integer
  ATTRIBUTE       Idle-Timeout            28      integer
  ATTRIBUTE       Termination-Action      29      integer
  ATTRIBUTE       Called-Station-Id       30      string
  ATTRIBUTE       Calling-Station-Id      31      string
  ATTRIBUTE       Acct-Status-Type        40      integer
  ATTRIBUTE       Acct-Delay-Time         41      integer
  ATTRIBUTE       Acct-Input-Octets       42      integer
  ATTRIBUTE       Acct-Output-Octets      43      integer
  ATTRIBUTE       Acct-Session-Id         44      string
  ATTRIBUTE       Acct-Authentic          45      integer
  ATTRIBUTE       Acct-Session-Time       46      integer
  ATTRIBUTE       Acct-Terminate-Cause    49      integer
  ATTRIBUTE       NAS-Port-Type           61      integer
  ATTRIBUTE       Port-Limit              62      integer



  5.  Proposals/unresolved issues

  There have been proposals to offer other flavors of database access
  through this library. It is hoped that the library will prove flexible
  enough to support this need. But such support is not currently the
  primary goal. In other words Cristian and Andrew will not be doing it
  in the short term.. so unless you want to volunteer, please don't hold
  your breath! :)


  6.  References

  7.  Acknowledgments

  This document was written by Cristian Gafton and Andrew G. Morgan.


  Thanks are also due to the following for their many helpful comments
  and suggestions: Nalin Dahyabhai, Torsten Duwe, Elliot Lee, Marek
  Michalkiewicz, Aleph One and Roland Schemers.


  More names to be added..