ASID Working Group                                            Y. Yaacovi
INTERNET-DRAFT                                                 Microsoft
                                                                 M. Wahl
                                                     Critical Angle Inc.
                                                               K. Settle
                                                               Microsoft
                                                             T. Genovese
                                                               Microsoft

Expires in six months from                               14 October 1996

Intended Category: Standards Track

                   Lightweight Directory Access Protocol:
                  Extensions for Dynamic Directory Services
                    <draft-ietf-asid-ldapv3ext-02.txt>
                    <draft-ietf-asid-ldapv3ext-03.txt>

1. Status of this Memo

   This document is an Internet-Draft.  Internet-Drafts are working
   documents of the Internet Engineering Task Force (IETF), its areas,
   and its working groups.  Note that other groups may also distribute
   working documents as Internet-Drafts.

   Internet-Drafts are draft documents valid for a maximum of six months
   and may be updated, replaced, or obsoleted by other documents at any
   time.  It is inappropriate to use Internet-Drafts as reference
   material or to cite them other than as "work in progress."

   To learn the current status of any Internet-Draft, please check the
   "1id-abstracts.txt" listing contained in the Internet-Drafts Shadow
   Directories on ds.internic.net (US East Coast), nic.nordu.net
   (Europe), ftp.isi.edu (US West Coast), or munnari.oz.au (Pacific Rim).

2.  Abstract

   This document defines the requirements for dynamic directory services
   and specifies the format of request and response extended operations
   for supporting client-server interoperation in a dynamic directories
   environment.

   The Lightweight Directory Access Protocol (LDAP) [1] supports
   lightweight access to static directory services, allowing relatively
   fast search and update access.  Static directory services store
   information about people that persists in its accuracy and value over
   a long period of time.

   Dynamic directory services are different in that they store
   information about people that only persists in its accuracy and value
   while people are online.
   when it is being periodically refreshed.  This information is stored
   as dynamic entries and/or dynamic attributes.  A typical use will be a
   client or a person that is either online - in which case it has an
   entry or attributes in the directory, or is offline - in which case
   its entry or attributes disappear from the directory.  Though the
   protocol operations and attributes used by dynamic directory services
   are similar to the ones used for static directory services, clients
   that are bound to a store dynamic information in the directory service need to periodically
   refresh their presence
   at the server this information, in order to keep directory prevent it from disappearing.
   If dynamic entries or attributes are not refreshed within a given
   timeout, they will be removed from getting stale in the
   presence of directory.  For example, this
   will happen if the client application crashes. that set them goes offline.

   In addition, static directories may include dynamic information, in
   the form of dynamic attributes in an otherwise static entry.  Such
   information needs to be refreshed periodically in the same manner.

   A flow control mechanism from the server is also described that allows
   a server to inform clients how often they should refresh their
   presence.

3. Requirements

   The protocol extensions must allow accessing dynamic directories information in a
   directory in a standard LDAP manner, to allow clients to access static
   and dynamic
   directories information in the same way.

   By definition, dynamic entries or dynamic attributes are not
   persistent and clients may go away gracefully or not.  The proposed
   extensions must offer a way for a server to tell if entries or
   attributes are still valid, and to do this in a way that will scale
   for a relatively large number of users. is scalable.
   There also must be a mechanism for clients to reestablish their entry
   or attributes with the server.

   Finally, to allow clients to broadly use the dynamic extensions, the
   extensions need to be registered as standard LDAP extended operations.

4. Description of Approach

   The Lightweight Directory Access Protocol (LDAP) [1] permits
   additional operation requests and responses to be added to the
   protocol.  The support for dynamic directories  This proposal takes advantage of these to establish a mechanism to support dynamic
   directories which contain dynamic information in a manner which
   is fully integrated with LDAP.

   The approach described in this proposal defines dynamic entries and
   dynamic attributes in order to allow implementing directories with
   dynamic information.  An implementation of dynamic directories, must
   be able to support both dynamic directory entries and dynamic
   attributes.

4.1. Dynamic Entries and the dynamicObject object class

   A dynamic entry is an object in the directory tree which has a
   time-to-live associated with it.  This time-to-live is set when the
   entry is created.  The time-to-live is automatically decremented, and
   when it expires the dynamic entry disappears.  By invoking the refresh
   extended operation (defined below) to re-set the time-to-live, a
   client can cause the entry to remain present a while longer.

   A dynamic entry is created by including the objectClass value given in
   section 6 in the list of attributes when adding an entry.  This method
   is subject to standard access control restrictions.

   The extended operation covered here, allows a client to refresh a
   dynamic entry by invoking at intervals refresh operations containing
   that entry's name.  Dynamic entries will be treated the same as
   non-dynamic entries when processing search, compare, add, delete,
   modify and modifydn operations.  However if clients stop sending
   refresh operations for an entry, then the server will automatically
   and without notification remove that entry from the directory.  This
   removal will be treated the same as if the entry had been deleted by
   an LDAP protocol operation.

   There is no way to change a static entry into a dynamic one and vice-
   versa.  If the client is using Modify with an objectClass of
   dynamicObject on a static entry, the server must return a service
   error either "objectClassModsProhibited" (if the server does not allow
   objectClass modifications at all) or "objectClassViolation" (if the
   server does allow objectClass modifications in general).

   A dynamic entry may be removed by the client using the delete
   operation.  This operation will be subject to access control
   restrictions.

   A non-dynamic entry cannot be added subordinate to a dynamic entry:
   the server must return an appropriate update or service error if this
   is attempted.

4.2. Dynamic Attributes

   A similar concept is dynamic attributes.  These attributes behave the
   same as ordinary user attributes to Search, Compare and Modify
   requests, except that if they time out they disappear from the entry
   in which they are located.  (Unlike the dynamicObject class, the entry
   itself does not disappear, and non-dynamic attributes are unaffected).
   Like dynamic entries, a static entry with dynamic attributes must have
   the entryTtl attribute which is described in section 6 below.

   Dynamic attributes do not necessarily introduce a new set of
   attributes in the schema.  Any static attribute in the schema can be
   made dynamic by including the ;dynamic attribute modifier with the
   typename.  It is allowed to introduce new dynamic attributes which are
   not useful in the static domain.  Such attributes will still require
   including the ;dynamic attribute modifier with the typename to define
   them as dynamic.

   A dynamic attribute may be a subtype of a non-dynamic attribute, ] but a
   non-dynamic attribute must not be a subtype of a dynamic attribute.
   Operational attributes and collective attributes must not be dynamic.
   All dynamic attributes are considered to be user attributes, however
   they are not guaranteed to be present in shadow copies of entries.

   A client may introduce dynamic attributes into an entry by using a
   Modify request to add them, or by including them in the attribute list
   of an Add Request.  If the attribute type is permitted in the entry
   then the dynamic attribute is also permitted.  The client would
   specify that the attribute is dynamic by including the tag ";dynamic"
   with the typename.  Dynamic values may be changed, and the attributes
   removed, by using the Modify request as normal.  If the entry is
   deleted, dynamic attributes disappear immediately along with all the
   non-dynamic.  A ;dynamic modifier on an attribute may not be used in a
   compare request, or in a search request (either in the filter or
   attributes requested for return).  This is in keeping with the
   philosophy that dynamic attributes be indistinguishable from static
   attributes as much as possible. In such a case, a server is expected
   to reject the request, returning 'invalidAttributeSyntax'.

   The granularity of a dynamic attribute is the entire attribute
   including all values.  Dynamic and static values may not be mixed
   within a single attribute.  For example, suppose an existing static
   entry had an attribute called "ipAddress" with the value "1.2.3.4".  A
   modification of the entry to add the value "5.4.3.2" to the attribute
   "ipAddress;dynamic" would fail with a "constraintViolation" error.

   The addition and modification of dynamic attributes are subject to
   schema and access control restrictions, as with non-dynamic
   attributes.  Thus unless the extensibleObject object class is present,
   generally an object class or content rule must be defined to permit
   those attributes to be present in an entry.  If their presence is
   controlled by an object class, then just as with non-dynamic
   attributes, the object class value must have already been added before
   the attributes are added.  The dynamicObject object class described in
   section 4.1 does not itself permit any particular dynamic attributes.

   Dynamic attributes in a particular entry are refreshed using the
   Refresh extended operation.  All of the entry's dynamic attributes are
   refreshed by a single refresh request.  The TTL given in the refresh
   response applies to all of the entry's dynamic attributes.  There is
   no way to refresh particular dynamic attributes within an entry at
   different times, or to have different TTLs apply to different dynamic
   attributes.

   If not refreshed, all dynamic attributes in an entry time out
   simultaneously.  All the attributes which are dynamic with all their
   values disappear atomically, as if the server had done a ModifyEntry
   specifying that all the dynamic types were to be deleted from that
   entry.  The client must not expect any dynamic attributes to be
   present in an entry after the time-to-live for that entry has reached
   zero.  However the attributes may not disappear immediately as servers
   may only process timing out attributes at intervals (e.g.  every few
   minutes).

   Note that if an object class definition references a dynamic attribute
   as a mandatory attribute, the entry will still time out, but a schema
   inconsistency will be present in that entry.  (The objectClass
   attribute and its values always remain since objectClass is not a
   dynamic attribute.) Thus it is strongly recommended that designers not
   specify dynamic attributes as anything other than optional attributes
   of auxiliary classes.

   Dynamic attributes may be used within dynamic entries (i.e., entries
   containing object class "dynamicObject", defined below), but since all
   of such an entry's attributes are implicitly dynamic, such use is
   superfluous.

5. Protocol Additions

5.1 Refresh Request

   Refresh is a protocol operation sent by a client to tell the server
   that the client is still alive and the dynamic directory entry or the
   dynamic attributes of the entry are still accurate and valuable.  The
   client sends a Refresh request periodically based on the value of the
   client refresh period (CRP).  The server can request that the client
   change this value.  As long as the server receives a Refresh request
   within the timeout period, the directory entry is guaranteed to
   persist on the server.  Clients implementers should be aware that
   since the intervening network between the client and server is
   unreliable, a Refresh request packet may be delayed or lost while in
   transit.  If this occurs, the entry or the dynamic attributes may
   disappear, and the client will need to detect this and re-add the
   entry or the attributes.

   A client may request this operation by transmitting an LDAP PDU
   containing an ExtendedRequest.  An LDAP ExtendedRequest is defined as
   follows:

         ExtendedRequest ::= [APPLICATION 23] SEQUENCE {
                 requestName             [0] LDAPOID,
                 requestValue            [1] OCTET STRING }

   The requestName field must be set to the string
   "1.3.6.1.4.1.1466.101.119.1".

   The requestValue field will contain as a value the DER-encoding of the
   following ASN.1 data type:

        SEQUENCE {
                entryName  [0] LDAPDN,
                requestTtl [1] INTEGER
        }

   The entryName field is the UTF-8 string representation of the name of
   the dynamic entry or the entry containing dynamic attributes [3].
   This entry must already exist.

   The requestTtl is a time in seconds (between 1 and 31557600) that the
   client requests that the entry or the dynamic attributes exists in the
   directory before being automatically removed.  Servers are not
   required to accept this value and might return a different TTL value
   to the client.  Clients must be able to use this server-dictated value
   as their CRP.

5.2 Refresh Response

   If a server implements this extension, then when the request is made
   it will return an LDAP PDU containing an ExtendedResponse.  An LDAP
   ExtendedResponse is defined as follows:

        ExtendedResponse ::= [APPLICATION 24] SEQUENCE {
                responseName            [0] LDAPOID OPTIONAL,
                response                [1] OCTET STRING OPTIONAL,
                standardResponse        [2] LDAPResult }

   The responseName field contains the same string as that present in the
   request.

   The response field will contain as a value the DER-encoding of the
   following ASN.1 data type:

        SEQUENCE {
                responseTtl [1] INTEGER
        }

   The responseTtl field is the time in seconds which the server chooses
   to have as the time-to-live field for that entry.  It must not be any
   smaller than that which the client requested, and it may be larger.
   Please note that for the case of a static entry with dynamic
   attributes, this time-to-live applies to all the dynamic attributes in
   this entry.

   If the operation was successful, the errorCode field in the
   standardResponse part of an ExtendedResponse will be set to success.

   In case of an error, the responseTtl field will have the value 0, and
   the errorCode field will contain an appropriate value, as follows: If
   the entry named by entryName could not be located, the errorCode field
   will contain "noSuchObject".  If the entry is not dynamic, or does not
   have dynamic attributes, the errorCode field will contain
   "objectClassViolation".  If the requester does not have permission to
   refresh the entry, the errorCode field will contain
   "insufficientAccessRights".  If the requestTtl field is too large, the
   errorCode field will contain "sizeLimitExceeded".

   If a server does not implement this extension, it will return an LDAP
   PDU containing an ExtendedResponse, which contains only the
   standardResponse element (the responseName and response elements will
   be absent).  The LDAPResult element will indicate the protocolError
   result code.

   This request is permitted to be invoked when LDAP is carried by a
   connectionless transport.

   When using a connection-oriented transport, there is no requirement
   that this operation be on the same particular connection as any other.
   A client may open multiple connections, or close and then reopen a
   connection.

6. Schema Additions

   All dynamic entries must have the dynamicObject value in their
   objectClass attribute. This object class is defined as follows
   (using the ObjectClassDescription notation of [2]):

   ( 1.3.6.1.4.1.1466.101.119.2 NAME 'dynamicObject'
     DESC 'This class, if present in an entry, indicates that this entry
           has a limited lifetime and may disappear automatically when
           its time-to-live has reached 0.  There are no mandatory
           attributes of this class, however if the client has not
           supplied a value for the entryTtl attribute, the server will
           provide one.'
     SUP top AUXILIARY )

   Furthermore, the dynamic entry or the static entry with dynamic
   attributes must have the following operational attribute.  It is
   described using the AttributeTypeDescription notation of [2]:

   ( 1.3.6.1.4.1.1466.101.119.3 NAME 'entryTtl'
     DESC 'This operational attribute is maintained by the server and
           appears to be present in every dynamic entry or an entry that
           contains dynamic attributes.  The attribute is not present
           when the entry does not contain the dynamicObject object class
           or dynamic attributes.  The value of this attribute is the
           time in seconds that the entry or the dynamic attributes in a
           static entry will continue to exist before disappearing from
           the directory.  In the absence of intervening refresh
           operations, the values returned by reading the attribute in
           two successive searches are guaranteed to be nonincreasing.
           The smallest permissible value is 0, indicating that the entry
           may disappear without warning.  The attribute is marked
           NO-USER-MODIFICATION since it may only be changed using the
           refresh operation.'
     SYNTAX 'Integer' SINGLE-VALUE NO-USER-MODIFICATION USAGE dSAOperation )

7. Client and Server Requirements

7.1 Client Requirements

   Once a dynamic entry has been created, clients are responsible for
   invoking the refresh extended operation, in order to keep that entry
   present in the directory.  The same holds true for keeping dynamic
   attributes present on a static entry.

   Clients must not expect that a dynamic entry will be present in the
   DIT after it has timed out, however it must not either require that
   the server remove the entry immediately (some servers may only process
   timing out entries at intervals).  If the client wishes to ensure the
   entry does not exist it should issue a RemoveRequest for that entry.
   Again, dynamic attributes behave the same: they will not be present in
   the DIT after they timed out, but the client should not require their
   immediate removal.

   Initially, a client needs to know how often it should send refresh
   requests to the server.  This value is defined as the CRP (Client
   Refresh Period) and is set by the server based on the entryTtl.  Since
   the AddRequest is left unchanged and is not modified in this proposal
   to return this value, a client must issue a Refresh extended operation
   immediately after an Add that created a dynamic entry.  The Refresh
   Response will return the CRP (in responseTtl) to the client.

   Clients must not issue the refresh request for dynamic entries which
   they have not created or for static entries to which they did not add
   dynamic attributes.  Please note that when a client refreshes a static
   entry to which it added dynamic attribute(s), it refreshes ALL the
   dynamic attributes in this entry, including ones added by other
   clients.  Also note that servers which allow anonymous clients to
   create and refresh dynamic entries and attributes will not be able to
   enforce the above.

   Clients should always be ready to handle the case in which their entry
   or dynamic attributes timed out.  In such a case, the Refresh
   operation will fail with an error code such as noSuchObject, and the
   client is expected to re-create its entry or re-add the dynamic
   attributes.

   Clients should be prepared to experience refresh operations failing
   with protocolError, even though the add and any previous refresh
   requests succeeded.  This might happen if a proxy between the client
   and the server goes down, and another proxy is used which does not
   support the Refresh extended operation.

7.2 Server Requirements

   Servers are responsible for removing dynamic entries and dynamic
   attributes when they time out.  Servers are not required to do this
   immediately.

   Servers must enforce the structural rules listed in above section 4.

   Servers must ensure that the operational attribute described in
   section 6 is present in dynamic entries or in static entries that
   contain dynamic attributes.

   Servers are permitted to check the authentication of the client
   invoking a refresh extended operation, and only permit the operation
   if it matches that of the client which created the dynamic entry or
   added dynamic attributes to this entry (please see a note about that
   in section 7.1).  Servers may permit anonymous users to refresh
   entries.

   Servers may require that a client which attempts to create a dynamic
   entry have a remove permission for that entry.

   Servers which implement this extension must have the OBJECT
   IDENTIFIERs, described above in section 5 for the request and
   response, present as values of the supportedExtension field in the
   root DSE.  They must also have as values in the attributeTypes and
   objectClasses attributes of their subschema subentries, the
   AttributeTypeDescription and ObjectClassDescription from section 6.

   An implementation of dynamic directories, must be able to support both
   dynamic directory entries and dynamic attributes.

8. Implementation issues

8.1 Storage of dynamic information

   Dynamic information is expected to change very often.  In addition,
   Refresh requests are expected to arrive at the server very often.
   Disk-based databases that static directory services often use are
   likely inappropriate for storing dynamic information.  We recommend
   that server implementations store dynamic entries in memory sufficient
   to avoid paging.  This is not a requirement.

   We expect LDAP servers to be able to store static and dynamic entries.
   If an LDAP server does not support dynamic entries, it should respond
   with an error code such as objectClassViolation.  Such a server might
   still support setting dynamic attributes on a static entry.

8.2 Client refresh behavior

   In some cases, the client might not get a Refresh response.  This may
   happen as a result of a server crash after receiving the Refresh
   request, the TCP/IP socket timing out in the connection case, or the
   UDP packet getting lost in the connection-less case.

   It is recommended that in such a case, the client will retry the
   Refresh operation immediately, and if this Refresh request does not
   get a response as well, it will resort to its original Refresh cycle,
   i.e.  send a Refresh request at its Client Refresh Period (CRP).

9. Localization

   The are no localization issues for this extended operation.

10. Security Considerations

   Security issues are not addressed in this document.  Please note,
   though, that anonymous clients are able to refresh entries which they
   did not create.

   Also, Care should be taken in making use of information obtained from
   directory servers that has been supplied by client, as it may now be
   out of date.  In many networks, for example, IP addresses are
   automatically assigned when a host connects to the network, and may be
   reassigned if that host later disconnects.  An IP address obtained
   from the directory may no longer be assigned to the host that placed
   the address in the directory.  This issue is not specific to LDAP or
   dynamic directories.

11. Acknowledgments

   Design ideas included in this document are based on those discussed in
   ASID and other IETF Working Groups.

12. Authors Addresses

       Yoram Yaacovi
       Microsoft
       One Microsoft way
       Redmond, WA 98052
       USA

       Phone:  +1 206-936-9629
       EMail:  yoramy@microsoft.com

       Mark Wahl
       Critical Angle Inc.
       4815 W. Braker Lane #502-385
       Austin, TX 78759
       USA

       EMail:  M.Wahl@critical-angle.com

       Kent Settle
       Microsoft
       One Microsoft way
       Redmond, WA 98052
       USA

       Phone:  +1 206-936-3027
       EMail:  kentse@microsoft.com

       Tony Genovese
       Microsoft
       One Microsoft way
       Redmond, WA 98052
       USA

       Phone:  +1 206-703-0852
       EMail:  tonyg@microsoft.com

13. Bibliography

   [1] M.Wahl, T.  Howes, S.  Kille, "Lightweight Directory Access
       Protocol (Version 3)".  INTERNET DRAFT
       <draft-ietf-asid-ldapv3-protocol-01.txt>

   [2] M.Wahl, A.Coulbeck, T.  Howes, S.  Kille, "Lightweight Directory
       Access Protocol Standard and Pilot Attributes".  INTERNET DRAFT
       <draft-ietf-asid-ldapv3-attributes-02.txt>

   [3] M.Wahl, S.Kille, "Lightweight Directory Access Protocol UTF-8
       String Representation of Distinguished Names".  INTERNET DRAFT
       <draft-ietf-asid-ldapv3-dn-00.txt>

   Expires on six months from the post date (see top).