Integrating Multiple Directory Services. Digital Common Directory Interface, OM, Version 2, Atom, TruCiuster, Integrated Directory Services
تبلیغات
تبلیغات
Integrating Multiple
Directory Services
I
Margaret
Olson
Laura
E. Holly
Colin
Strutt
The Integrated Directory Services (IDS) i nfra structure i m plements a directory-service independent i nterface. The IDS i nfrastructure is used by applications that store and retrieve information a bout resources in environments with either m ultiple d i rectory services or one of several di rectory services. The I DS interface isolates users and appl ication writers from the unique requirements of different directory services by providing a view of a single, logi cal d irectory service through a simple federa tion mechanism. To retrieve resources from the logical directory, IDS determ ines its phys ical location and converts the resource from a directory-specific to a canonical format.
Extensible schema tables represent the canon ical format for each resource and allow I DS to represent resources created using both the I DS interfaces and the directory-specific interfaces.
Digital has developed the Integrated Directory
Services
( I D S ) tec h nology ro provide a mechanism for integratjng m u l tiple d irectory services into a single sys tem . In this paper, we examine the development of the
I DS i n fi·astructure . We begin by ruscussing the prob lems faced by network directory appl ications. Next we describe our design goals, the
IDS i n frastructure, and our initial implemen tation on the
I'AT HWO lU(S prod uct. We concl ude with a brief discussion of plans br future development.
Di rectory Support in Multiple Environments
Al though rurectory services are a powerfu l mechan ism t(Jr distributing and accessing certain kjnds ofinfonna tion, relatively tew Jpplications choose to use them .
Digital's
PATHW O RKS application was in need of a directory for printers :md fi le shares.
�1 net\vork operating system
( N OS
PATH WORKS is
) integration product that gives users access to both M icrosoft's LA.!'\l
Manager and Novell's NetWare file and print shares. As we studied how to incorporate directory support into
PATHWOIU(S , we came to a better understanding of the problems faced by directory applications in general .
Networks are growing rapidly, as are the amount and kj ncf of information that can be accessed through the net\vork. We were certain that fu tu re network appl ication products wou ld have an even greater need tor a directory, and therefore a general sol ution was needed. We then set out to design a system that would remove the barriers to di rectory service <�pplication usage and deployment. We resolved the tension bet\veen the product dead line and the time req u i red ro implement rhe ge neral solution by design ing a complete solution and i m p lementi ng what was neces sary to prove the design and to meet the i m medi<�te needs of the
PAT HWO RKS product.
Existing Directory Services
There are a n u m ber of general -purpose directory ser vices. Some of the more frlmiliar include
Nove l l 's NetWare Directory Service
( N O S
X.500,
)
, the Cel l
Directory Service
( C DS
)
, and Banyan Systems'
46 Digital Ted1nid jound Vul. 8 No. 1 1 996
Strl:l:tT�1lk.
1 1 ln the past, dirl:ctorv sl:n·icl:s Wl:IT in rl:l ati\Tiv l i m i ted usc because most d irl:ctorv sn,·iccs wnl: til:d to l:ithcr an operating S\'Stcm or a transport or both.
In addition, d irector\' services \\'ere con tKctl:d to a m u l titude of application programming interbcl:s (APis) that were incomp<Hible and difticult to usc More recently, directory services h;we been tied to network operating systems or applications, rather than to host opeL1ting systems or transports. If any thing, the nu mber of"standard " A Pis Ius grown.
I n L1rge networks, this complexity has resu l ted in the proliferation of directories, often contJining over lapping i n �{>nllation. This ma kes the network man ager's job d i fticu lr, which in w rn creates tTsistancc to d i rector\' appl ications_ At the sanK rime, network and
N O S tcchnolog-v hJs developed to a poi nt whne :111 l:ver-incrctsing :1mounr of information is being shared
011 different machines. To give a simple example,
�1lmosr every server :1t Digital's Littleton site hJs a con nection to the h i gh -volume printer in thl: copy center, with a d i ffnent name on every server.
A directory wou ld simplify users' access to this single physicJ I resou rce by presenting a single name �()r the printer, i f onlv the application writer could �igurc out which dirccton· sen·ice to use Jnd how to use it.
Other Approaches
As discussed later i n the Design of the
�llld
Service Providers section,
I DS
I DS
hJmework defines both an APJ and �1 service provider inter��lCe. Support tc>r �1ny direc tory scn·ice
Clll be provided by writing a service prm·ider module. lYlicrosoft's
O L E
Direcrory Su·vices
(OLE
DS)
takes J similar approach to
I DS ,
with a more limited initiJI implementation .' A l though the current
I DS
implementation runs under Microsoft Windows, it w:Js designed to port to other systems.
OLE DS
depends on tCJtures of the vVi ndows operating systems.
The X/Open federated Naming ( XFN ) specificJtion was not complete at the time we were designing
I DS ,
�llld it did not include either :1 service provider interbce or
:1 rdcrence implementJtion 6 We did ex<1mine the
XFN d raft and designed the
I DS interfJce to be com p�niblc with
XFN , with a view toward su pporting the
XfN API in the future. Supporting the XFN interbces on top of
I DS wou ld be a rehtivelv srraightrc>rw:Jrd task, �md we llJve considered doing this_
The PATH WORKS Application
I n rhe
NOS environment, each
NOS h�1s irs own directon· or pseudo-directory_ NetW:Jre version 3 implcmnns the Binderv; NervVare 4 implements
N DS .7 TIK \'Jrious implementations of Microsott's
LAN Manager protocols provide a ''irtual directory b�1sed on i n tcm11ation maintained bv irs dom:1in con tro l lers.
In
:1 mu ltiple
NOS eil\'ironment, the user is presented with m u l tiple i n�(mn:Jtion sources ti·om the m ultiple directories. E\'Cil worse, the user mav be taced with multiple in�cm11ation sources even in a sin gle
N O S environment, since there ma\' be m ultiple
NetWare Binderies or LAN Manager domains.
Multiple
N OS
environments do not, in and of them selves, cause complexity and confusion . Problems arise when people within �1 single environment want to sh<ll'C resources across multiple environments. For example, consider a common local area network ( LA N ) configu ration where NetW:Jrl: is inst:Jikd on the c l ients :111d servers tor one departmellt and Microsoft's LAN
Manager ( contained within products such as
1'v1icrosott's Windows t(>r Workgroups, Windows 9 5 , a n d Windows
N'T'
operati ng systems, or the LAN
Server product ti·om l nternation:JI Business JV!Jchines
Corporation ) is insr:: d lcd on the clients and servers tor another department. I f each department's resources, users, and administration personnel Jre kept distinct, there is no problem . However, :111)' desire to al low users to share resources between d e partment s
, or to have common �ld min istrJtion over the departments i ntroduces administrative and user pro bl e ms.
I f a printer is to be shared by the two departments, it must be administered twice : once in the NetWare environ ment and once in the LAN Manager environ ment.
Users in the two dcp:1rrments use d i fferent nJmcs t(H· the same primer. LJter
NOS
i mplementations, such as
DigitJI 's
PATHWO RKS
,·ersion 5 . 0 or the networking sofuv:1re b u i l t into Microso�i:'s Windows 95 that pro vides support tor m u ltiple
N OS
protocols, do nothing to manage the mu l tiplicity of names for the same net work resource .
As we were contemplating the set of capabilities we needed to design tc>r the next generation of
PATHWOlU<.S
c l ient products, we reJiized that solv ing the connectivity problem i mplied i n a m u l tiple
NOS
environment w:1s not enough . User access and administrJtor control of
N O S
resources needed to be considerablv simpler.
As we looked Jt the problems i n larger networks, we saw the need t(>r the abilitv to pr o vid e more sophis ticated means to locJte
N OS
resources. TypicJIIv,
N OS
c l ient sott\\'are provides the means to bro\\ se the network to locate a resource . Hown·er, browsing requires the user ro know the location of the resource, specifically the name of the server, and to be able to choose the resource on rhc server by recognizing something about thl: resou rce n:1mc or J resource description provided by the :1d ministrator. What w:1s needed was a design thJt allows a user to search, Js we l l as bro\\'SC, t(>r
�l resource bJsed on \'Jrious attrib utes describing the resource .
Fin:11Jv, existing
NOS ell\'ironments have a fairly lim ited ,·ie\\' of the set of resources tllJt can be referenced .
Digir,,[ Tcdmid J uunnl
Vol . 8 No. 1 1 990
47
48
Both NetVVMe and 1·arious
LAN
J'vLm�1ger i m plemen t:nions pro1·ide support f<>r p ri mers and tile sh�1res .
We wamed to be able to extend the t\'pes of resou rces th�1t could be refere nced and lll<111Jged f[·om the new directory capabi lity that we were designing.
Thus we em barked on a design r()r the facility we initially cal.led
I DS , for lntegr.ned Directory Serl'iccs.
The
PA'rJ-fWO R.KS version 6.0 implementation was cventll�< l l v cal led Directory Assistant. We refer to this rcclmology <1S
I DS throughout this paper.
Design Goals
As we looked at the rcquircmcms of the PATHVVO RKS product, we found rh:n m:1nv of those req uiremenrs wJs integrated into the
PATH WO RKS applications and tool sets.
PATHWO RKS required the :1bility to
•
•
•
Gi1'C a single name to resources thJt em be accessed lw means of m u l tiple sen'Crs or protocols l ns u !Jte end users ti·om ch:mges in the w�1\' resou rces are a l locJted :1mong the servers
M:111:1ge resources in a n
NOS-i ndependent manner
\'Ve coul d not simply pick :1 di r ectory service and integrate it i nto
PATH VVO RKS, beclllse we cou ld not require that all c ustomers dcplov J parricubr directon· scn·ice Jt their site . The
PAT I:-I WO !\KS product is both
N O S and transport - i ndependent ; introducing such :1 dependence was u n:Jcccpt:< b l c . We q llickh· rc:�l ized thJt these were the req ui remenrs that kept ma111' other :1pplications from using di n:ctot'l' sen ices.
Our assu mption \\'JS that
111 JIW nct\\'Ork applica tions wou ld usc d irecton· scn·ices if thev could, bur that kw of them could assun1c or req u i re a p:1rtic u lJr d ir e cto r y service. Working fi·om that assu mption, we selected the follo w i n g d es ign requirements for
I DS :
•
•
•
•
•
•
•
Directory service independence
Abilitv to access existing d,m
Abi lirv to join disp:�r:Hc namesp:�ces i nto a single, logical namespace
Rcmm·:�l of barriers to successfu l dcplovmem of
:1 wide area network (WAN ) d i recton·
Abi lity to hide directory nJme syntax
Support of search
Support of application-specific directory entries
Directory Service Independence
Customers must be :�ble to choose the dirccton· sen icc in which the1· store resource i n formation . Some cus tomers ha1 c a prekrred director\' scn·ice, \\'hich the\'
11·:m t to cont i n u e to usc. Other customers, who arc not using a particu!Jr d i rectorv sen ice, prefer that Digital pn>�·ides the directorv sen·icc . I n :1 k11· cases, a cus tomer m ight wish or even need to store intonnation about d i tkrem resources in different directory sen·ices.
Ability to Access Existing Data
A great deal of i n f(mn�nion currently exists
1 11
�lppl ication-specific directorv services and in
NOS speci fic d i rectory scn·ices. A relarivelv large n u m ber of applications a lso usc the native i n terbces to store i n f(m11Jtion in the
N OS d irectories. Al lo\\'ing users to :Kccss this i n form:�tion d irect!\' through
I D S
\\'JS critical . We expressh· II'Jnted to :woid the need to d upliutc di recron· intorm�1tion in separate, incomp:�t i blc svsrems.
Ability to Join Disparate Namespaces into a Single,
Logical Namespace
Manv directory ser\'ices arc aimed Jt a specific applica tion or
J set of appl ic:nions. for example, c u rrent
X . SOO dcplovment s contain mostll' people i n form:J rion such as names, p hone n u m bers, and e lectronic mail :�dd resses. ( Note:
X . S OO is an exrrcmclv flexible di recrorv sen·ice that cJn be used to store almost an1· kind of' i n f(mnarion, bur f(>r historical reasons most deployments contain people i n f(m11ation . )
NOS d i rec tories contain i n t(xnl:ltion :�bou t
NOS resou rces such
:1s printers.
Conseq uently, many user environments hJI'e mu ltiple d irectory sen·ices, c1eh of which con rains critical business intonn:nio n . To access this exist ing d:na and present it to the user in
J meaningfu l 11·:11', these m u l tiple d i recron· n::unespaces must be joined i nto a single, logic:�l namespace .
Removal of Barriers to Successful Deployment of a WAN Directory
H ier:1rchicJI d i rectory services gener:�l l\' req u i re thJt the naming h ierarchy be designed berclrc the directory is deployed . Since t he hicr:1rchy consists of names, and names are sensitii'C and poliriul entities, this em be an e x tremely d i fficu l t task. Oq:�anizations also change m·er time, fu rther c o mpli c
Jti n g the problem of designing a name h ierarclw.'
Org:�nizations that successfu ll y deploy d i rectorv ser vices do so fi·om the bottom u p . The
NOS d irectories
:1rc deployed precisely because thev :woid the prob lems i n herent in a n:�me hier�1rclll'. An administrator c1n set up a Novell 3 .x Bindery t(lr a local organization without worrving about how the name o f one group rel:�tes to the names of a l l the other groups. The downside to the
NOS d irecto ries is that they have
,, l i m i ted :1bi l i t1· to scale be10nd a
LAN .
With
I DS ,
II'C
\\':Jilted to pro1·id e '' fr:�mework rh:1t wou l d gro\\· \\'ith the user's em·ironmenr. A user could stJrt ll'ith a loc1l d i rccrorv b u r incorporate th
:t t dircctorv imo
Jn enter prise or global di rectory "'hen the time \\'JS appropri
Jte, without affecting the end users or the applications.
Vol . H No. 1 1 9\1(>
Ability to Hide Directory Name Syntax
The syntax of the nJmcs in hicrJrchical directory ser vices varies nor only fi·om one directory service to another, bur in some c1ses fi·om one implementation of a single di recrorv service to another. The syntax for
Domain Name Svsrem names is ordered the same as a postal mail address, thJt is, fi·om the most-specific componcnt.
" · 111 For example, a machine at Digital might be bigAiplu.digit�1l .com. The X . SOO name order is usu:�lly ( d epend ing on the impleme ntation ) the reve rse. The corresponding X . SOO name might be: c = us;o= Digital ;cn = bigAipha. Particularh' in the
X . 500 case, difkrent svstems �md �1pplic:�tions also accept difrcrent sep:-trJtor cluracters.
Together, the lOS designers hJve much experience with a nu m ber of directorv services and their n:�me synt:�xes. Users and applications developers alike have been quick to poi1H out the problems with directory
Jumcs. These names :�rc cumbersome, conh1si ng, or just plain inconvenienr to type . The separator charac ters within
:-� directory n:�me may have special mean ings on some operating systems.
Because of these limitations, we decided that a name syntax specific to I DS would detract fi·om the value of the solution. An applicnion using I DS may choose to prescnr its own syntax, one that is suitable to i rs partic u lar environmenr Jnd prekrences. The AP I takes the object name and the context, as described in the
Contexts section. The service provider module uses these ro construct the n:1me in the native name synta x .
Support o f Search
Users need to locate resources in a nu mber of ways.
The most tamil i:�r met hod is to locate resources b\' knoll'ing the ir n:�mc; this is often reterred to as a white pages lookup, n�1med :�Iter the printed U .S. telephone direcrory of �1lph:1beticallv ordered names.
Searching for resources b:�scd upon information about the resources is referred to :�s �l vellow pages looku p, named after the printed U.S. telephone directory organ ized bv business category. To support vellow pages lookup, resources must be retrievable from the di rectory service based on their attri butes. For a printer, this might include the type of printer, the loca tion of the pri nter, whether it supports color or not, who is responsi ble f(.>r maintaining the printer, and other i n form:�tion. IDS needed to support both yellow pages and white p:1ges lookups.
Support of Application-specific Directory Entries
We saw a need to su pport two kinds of extensibi lity: the abi lity f(>r �111 application to crc:�te new kinds of d irectory entries, :�nd the :�bility for a customer ro add attri butes or other descri ptive i n f(xmation to the di rectorv entries created bv PATHWOR.KS or other applications. By providing appl iorions with the capa bil i ty to create new kinds of directory entries, the I DS designers al lowed IDS to be used by any application, regardless of i ts requ irements. By allowing the addi tion of attributes to existing directory entries, we allowed customers to easi ly :�dd i n formation that is speci fic to their organiz:�tion to application objects.
For example, a customer might add J speci fic cod e, such as an asset identification tag, to all printer direc tory entries.
Design of the IDS Fra mework and Service
Providers
I DS is an object- based system th:�t consists of a frame work and a set of service providers. For clarity, we fi.Jr ther divided the ti-amework into an A PI and a service provider interface ( S P T ) . The API consists of a su bset of the framework's objects and their public virtual methods. The SPI is a generalized, di rectory-service independent i ntertace (descri bed in detail later in this section ) . The SPI objects define the abstract interface to the directory service . We use the term
service prouiderto
reter to any directory service that provides
I DS storage. The service providers in teract with the framework through the SPI .
Framework
The fi·amework performs three major fu nctions:
•
•
•
It specities the IDS director y independent opera tions.
It
dispatches oper:�tions to directory-specific mod ules for execution.
It verifies that all I DS objects and operations do not violate the I DS schema.
F igure l ill ustrates the structu re of IDS. When an app�cation makes an API cal l, the fra mework examines tbe name information and calls the appropriate service provider. The service provider then ma kes the call to the appropriate native directory service client. When the d i rectory client returns the results, the service provider converts the resu lts i nto the I DS canonical form. The design supports ju nctions fl·om one d irec tory service to another, in that the result retu rned to the framework by the service provider may be only a reference to an object in another di rectory service.
The abstract interbce to the directory service ensures that I DS provides applica6ons with a consistent level offunctiona�ty without regard to which directory service a customer has in his or her environment.
Because the words "object" and "object class" are overloaded and overused in the industry, we define the words "resource" and "resou rce class" to denote objects represented in I OS. A
resource
is J di rectory
Digiral Teclmi(al Journal
Vol .
8 No. I 1996
49
IDS
APPL ICATION PROG R A M M I N G INTER FACE
SERVICE PROVIDER INTERFACE
'--
SERVICE
PROVIDER
SERVICE
PROV I D E R
SERVICE
PROV I D E R
I-
NATIVE
D I R ECTORY
CLI ENT
FRAMEWORK
NATIVE
D I R ECTORY
CLIENT
NATIVE
D I R ECTORY
CLIENT
I
�
�
�
NATIVE
DIR ECTORY
SERVER
NATIVE
DIRECTORY
S E R V E R
NATIVE
D I RECTORY
S E R V E R
Figure 1
Structure of the I ntegrated Directory Services e n try; it is a directory service object that represents some network object. A resource class is the defi nition of that type of d irectory entry. For example, the direc tory entr y that describes a specific printer is an I DS resource, and the I DS class that describes every printer entry is a resource class.
The framework provides extensibility by defining
C++ object classes that allow for the creation and manipulation of resources, attributes, and attribute val ues i n a type- independent manner. The type inde pendence al lows both appl ications and the ti·amework itself to manipulate I D S attributes and attribute values without knowing their types. As long as the new types are built on top of existing I DS system types, applica tion writers may ddine new IDS types without mod i fYi n g the service providers.
The fi·amework d ispatches d irectory operations to the appropriate service provider and maintains overall system state and integrity. Tt maintains a l ist of the service providers that are currently available a nd shows the errors encountered in any fai led loads.
This allows the svstem to continue to operate, albeit in a degraded st
� te, even though one of the service providers may be malfunctioning.
Before we d iscuss the design of the SPI, we describe tl1e framework's objects.
IDS Entry
The fundamental IDS object is the canon i c a l representation o f a d irectory entry, t h e IDS en try.
The I DS entry is a n abstract object. To create a resource cl ass, applications define a resource type and derive it from the IDS entry. IDS entry objects are cre ated and manipul ated through the API and translated into the appropriate native directory format by the ser vice providers. Derivatives of the IDS entry may define
J.d d i tional methods, but they may not override the
I DS entr y methods. The IDS entry methods are part of the framework.
The I DS entry methods tall i n to one of two otegories: those wh.ich manipulate the attributes and values contai ned in the I DS entry in a type-indepen dent manner, a nd those which perform operations on the directorv. Each I D S entry, each attribute, and each attri bute val
�
1 e contains a type. For convenience, deriv atives of the I DS entry may define additional methods that manipulate certain attributes or values di rectly.
For example, a derivation tl1at defines a printer might define a method to set rhe description attribute. The implementation of this method would cal l the general
I DS entry attri bute and value manipu lation method to set the value of the appropriate attribute.
As shown i n Figure 2, the IDS entry contains identi f),ing information and the attributes and attribute val ues that describe the resource. The context identi fies the service provider that performs directory opera tions on this entry and the location within that directory service in vvhich this entry is stored . The resource type detines the kind of resource that this entry represents. The resource name is the name by which applications and users refer to the entry.
The attributes of the entry are contained in a set.
Each attri bute in turn contains the vJiue or list of val ues associated witll the attribute.
Contexts
The context is an object that uniquely iden tif-ies a particular location in a particular namespace .
The IDS context i s very similar i n concept to the XFl'i context." All contexts contain the type i de ntifier for the directory service and an internal name. The type identi fier is used by the fDS framework to dispatch operations to the appropriate service provider. The internal name is the location within the directory ser vice described by this conrexr. The internal name is represented in the native syntax of the underlying di rectory service. The service provider is responsible tor setting and maintaining this internal name. ( See
Figure 2 . )
Attributes and Attribute Val ues
The type of an
::mribute defines the data type of its va lue or values.
The attribute value object is a canonical representation of an acmal attri bute value. The attribute val ue object de f-ines a set of methods tor accessi ng and manipu lat ing val ues. For each data type supported in IDS, there is a corresponding attribute value derivation in the
50
Digiral Tc-dmical
Jounul
Vol .
8
No.
I 1996
I
RESOURCE TYPE
IDS_PRINTER
SERVICE PROV I DER TYPE: LDAP
CONTEXT
LOCATION WITHIN SP: o=dec;ou=lkg
I
RESOURCE NAME
I
NIST GROUP PRINTER
ATTRIBUTE SET
ATTRI BUTE N
ATTRIBUTE
2
I
J
ATTRIBUTE
1
I
ATTRI BUTE TYPE
I
IDS-ATTR-MAINTAINER
ATTRIBUTE VALUE LIST
I
ATTRIBUTE VALUE DATA TYPE
I
ATTR IBUTE VALUE t
I
ATTR IBUTE VALUE DATA TYPE
I
ATTR I BUTE VALUE t
I
ATTR IBUTE VALUE DATA TYPE
I
ATTRIBUTE VALUE
I
I
I
I l i
I
DS-TYPE
I
JANE DO
-STRING
E
Figure 2
IDS
Entry
IDS framework . This allows applications, and the I DS framework itself, to manipulate attribute values with out knowing their types. The service providers, on the other hand , use the type information ro translate from the I DS data formats to their native data tonnats .
Types
To allow customers and third parties to identi�' their own IDS resources, the I DS type mechanism must uniquely identi�' objects. The two identifiers we considered using were universJI unique identifiers
(UU IDs) as defined by the Open Sofuvare Foundation
D istributed Computing Environment ( OSF DCE) and object identi tlers ( OIDs) JS defi ned by the open sys tems i nterconnection ( OS I ) standards.
" · ' 2
Some direc tory services identity attributes with OIOs, whi le others use UUIDs. For appl ications defin ing new resources, we wanted to avoid the necessity to obtain both an
OlD and a UUID . It is possible to encode a U U I D in an oro, but the reverse is not true.
We cou ld encode a UUID in an OlD by registering an OlD prefix. The prefix would indicate that the sequ ence after the prefix was a U U I D . U U I Ds are fixed-length structures generated from time sta mps and Ethernet addresses, and therefore arbitrary infor mation such as
:111
OlD cannot be encoded in them.
UUIDs are also e::tsier tor application writers to gener ate because nu merous systems ship with tools to generate them.
Certain directory services, for example
X . SOO, have external type definitions tor the directory entries. lt is possible to define a generic entry and then map arbitrary values inro that entry, but I DS entries would not be meaningfltl when viewed with the native di rec tory management tools. We fel t that this was unac ceptable, because it would make the management of
I DS entries in the namespace much more difficult.
Some systems use U U ! Ds to represent the type infor mation. We chose to use UU !Ds si nce they are both easy to generate and can be used in both U U I O and
OlD class definition systems. The use of OIDs wou ld require U U I Ds to be generated for U U I D -based systems and mappings to be maintained .
Digital
TcchniL·al Journal V(ll 8 Nu. 1 l996
5 1
Com m unities
An I DS com nH111irv i s both an adminis n·ativc grou pi ng mechanism and ,1 logic:�! loc
1 tion for
IDS resou rces . When people imcr�Kr wirh the I DS svs rem, rhev sec a com m u n i rv as the org�mizing principle.
The �1dminisrraror contro l s rl1e bou ndaries �111d mem bership of :111 IDS comi11UI1in·. T1·pic1llv, <1 com mun in· represents either a p�1rticuL1r locnion such as a b u i l d ing or a hmcrional grou ping such a s :1
\\ Ork group.
I n iri�1lly, \\'C co nsidered
:1 su p crco nrnt ro
j
o in multi ple directories i nro
:1 single logiol di rcctorv. This supercontnt wou ld ha1·c cont�1incd mul tipl e contexts, one
r(n
c�1eh n•pc of resource su pported bv I DS . We e1
· enrual lv subsu med the supcrconrot i mo a comrnu n i n• :111d oiled it a resource COI1tor l ist. An IDS com
mun in• is ston:d as �1 special object in rhe d i rectorv.
F.ach comm unity's resource comexr l ist describes rhc directories that 111�1kc up the com munirv. The resource context list is rhc kdcr:�rion mechanism by which I DS determi nes where resources of each type :�re stored .
Each entry in the resource context list is �1 pair of resource type :�nd context. As users and a pplic a t ions oper:�rc on entries in
:1 com munity, rhe I DS ti
· amework
(through
IDS enrry and communit\' methods) inspects the resource type and the com l nu nin· to
d
ercnni ne the context. Figure 3 i l l ustrates an IDS comnu1 11in•.
One of the probl ems we anticipated w�1s th�H large organi zations ll'ould nat u rallv re nd to han: 11
1 a nv IDS com muni ties: How wou ld r h c user idemi t-\· these; We considered Jn additionJI h icr:�rchv in
11 hich com m u ni ties ll'ou ld be mem bers o f other com m u n ities. Our usabi l i n· consultants cmpb
;
1siz e d that users should nor have to broll'se a hicrJrcll\' ro �Kcess resources . I n response, we developed the concepts of t h e loca l ami
. atcd \\'ith rhe machi ne J user is curremlv using
it represents a plwsical location . The home com m u 1
1 it)
' is the one with which the user is
J ssociat c d or belongs
. vVe e nv
i
sioned rh:�t the home com mu ni ty wou ld be the s:1me as the local communitv �1r the user's normal place of work, b u t thet-c is t10 req ui rement inherent in the design th:Jt things be orgJ n i zcd this W�1y. ror example, if a user is associ:ncd with the com m u n i rv at her work site and the nuc hinc she uses is
:1 l so Joc1tcd at that 1vork site, both her Joc1l communi ty and
COMMUN ITY
DEFAULT
CONTEXT --------
RESOURCE
CONTEXT
CONTEXT
RESOURCE CONTEXT LIST
OBJECT
TYPE
Svc Provider Type = F7801 DB7-F675-1 1 CD-ABC2-08002B1 87D1 A (ODBC)
External Name = IDS_Group Community lnlernal Name
=
E \\tuxedo\idsodbc\idsdbdir.mdb
Svc Provider Private = NULL
Svc Provider Type = EFF4B840-EC52-1 1 CD-9E5 E-08002BBA95CA (CDS)
External Name = ids_cell.lkg.dec.com
Internal Name = ids_cel l.lkg.dec.com
Svc Provider Private = NULL
Svc Provider Type = C723E850-A
1
A6-1 OAB-A699-08002B361 FC1 (LDAP)
External Name = c=us: o=dec;ou=IDS_Group Community
Internal Name = c=us:o=dec;ou =ID S_Group Community
Svc Provider Privale = YUMMY, 386. TCP/IP
52
KEY: c::::=J
COMMUNITY c::::=J
RESOURCE CONTEXT LIST
-
RESOURCE CONTEXT c=J
OBJECT TYPE
CONTEXT
Figure 3 l DS Com m u n i tv
\'ol. X :--:o I 1 996
her home comnHmiry rqm:senr this \\'Ork sire. If this user works at another work site and uses a different m:1.chine, her home community remains the s:1me, but her loc;1l comtnunity rdlecrs the community wlH.:re the new m:1chine resides. The concepts of local and home communities do nor red uce the nu mber of com munities, but they do provide a d irect method bv
\\'hich users can access the comm u nities that contain the resources the\' most hTquentlv use . The local :md home commu nities are a con,·en ience; users and :lppli cnions are in no wav restricted to those communities.
Search Support
Searching is handled by the search object. The search object contains a commu nity ( or list of communities ) , :1 resource tvpe, and Jn attribute tiltcr. The attri bute til tcr supports both equalirv and comp:1rison matching of attribute val ues and allows c:1llcrs to construct complex requests by conc:�ten;Hing comparisons together in a series of Boole:�n opera tions. For example, a cal ler cou ld construct a tilrcr tlut returned all pri mer objects that ( ( ( are locJtcd on F loor2 ) OR ( are located on Floor3 ) ) A N D (sup port color printing) ) . Combi ned with the lou I :md home commu nity supporr, ti l tcrs allow applications
:md users to express ideas such JS "print this at the closest printer that supports color, t\\'o-sided pri nting,
:�nd then transmit it to ;l n\' bcsi mile machine in 111\' home conHnu nirv. ''
The carch object's ddiult tiltcr returns all objects of the resource type in the locJI community. The search object resolves the community to a context :md passes it to the service provider. The service provider con structs a l ist of matching I I
)5 cntrv objects to return to the user. In I DS, the search object supports browsing.
The search object has mctl10ds that displ:�y a diJiog and construct tilrers bJsed on user input. vVhcn design ing the S\'Stem , we dcb:Jted whether it was bet ter f()r the se:1rc h object to cont<lin both the fil ter and rhe search d ialogs or whether the fi lter constru ction belonged in the IDS en try. We chose to keep the search dialogs separate h·om the I DS entry. Experience with i mplementing resources derived ti·orn the I DS enrn· h:�s shown this to be an error. Currently it is ncc css;ll'\' to derive ti·om two objects, I DS entrv and the sc1rch object, to implement a resource that has a resource-specific search dialog. We will be modirving the search and I DS entry objects so that the construc tion of the ti lters and the di:�log that constructs the ti ltcrs are I DS e ntry methods.
Schema
The ser\'icc prm·idcrs transbte between the n:�ti,·e directorv object and the I DS en tn·. I n genera l , directory sen·ice entries arc n o t self-describing. I n existing directory services, either J schema or the applic:�tion is expected to know the d irectory-spccitic f(mnat of the data . The btter is more common than the former, ,md in Jll\' case the schema methodologies
Jre unique to uch directOr\' service .
From the poinr of \'iew of the native dircctorv ser vice, IDS is the appl ic:�tion. To properly convert the data, the se rvice providers must knmv what it is. The service pro\'idcrs usc the schema to determine the cor rect attribute <1 1ld \'J iue types to use when constructing the
I DS entrv of;1 particular t\'pc .
The schema describes resource tvpcs, Jttri bu te tvpes, and attri bute \'alue data types. Logic:�l lv, the schema is a set of t:�blcs, one for each service provider, which maps the native nJme or type to the I DS name or type. These tables are read by the I DS schema com ponent when I DS is initialized. Bee:� use these tables are external to the system , thev can be modified by users or applicJtions.
There is one l i m i tation on the exte nsion of the schema: Ne\\' Jttribute and resou rce rvpcs can be ddined, but thev must be composed fi'om the prede fi ned IDS :lttri bute \'alue types that the scr\'ice providers can support. The service providers wou ld have to be mod i ticd to su pport Jddition�l attribu te va lue data types. This limitation is not as severe JS it
:lt first appears. A rich set of data types is ddined in the existi ng dirccrorv sen·ices, and a relarivclv sm:\11 set is in common us:1gc. Bv defining the I DS data types ro encompass the set of Lhta rvpes ddined lw existing di rectory services, we have reduced this l i mir:� rion to
J theoretical rather than a practical problem .
As a conseq uence of the use of schem:1, applications must specit)' rile resource type tor Jny I DS operation .
This is a limitation that in principle docs not exist in other directory svsrems. After some consideration, we concl uded that kw usefu l operations can be performed on an object whose type is un known . To perfcm11 an operation on objects of all rvpes, the schema c:�n be i nterrogated t()J' the l ist of all supporred I DS object types, and the opcr:�rion is then iterated over each type.
The System Object
The system object lo�1ds and initializes the service p roviders. On in iti:�lizarion, the svstem object constr ucts a list of the avai i:Jble service providers ti·om those ddined in a local configurJtion file.
The svstem object constructs and maint:1ins the l ist of known communities. The system object obtains this list using the following mechanisms:
•
•
•
Inspect a well- known location (if one exists) to see i f it cont:� i ns a cache of known comm unities.
For c:�ch sen·ice prm·ider, call the discmn method to ask the scn·icc provid er tor its list of known communi tics.
I f the system object is initializing for the tirsr rime, prompt the user to create a community.
Digir.1l T<:dmiul
)ourn;ll
Vol .
X No. I 1 996 5 3
Application Programming Interface
As mentioned previously, we divided the framework i nto an API and a service provider interface (SPI ) . The
APT consists of the search object methods, the IDS entry methods, the attribute object :�nd value object methods, and the system object methods necess:�ry to access communities.
Service Provider Interface
The SPI specifies the interface bet\veen the I DS frame work and the n ative d irectory services. I r defi nes the semantics for aU operations that may be performed on
I DS inf-ormation regardless of which directory service stores the i n f-ormation . The SPI effectively insu lates both the I DS framework and the I DS applications from the unique syntax and requirements of d i fferent di rectory services.
A directory-specific modu le, cal led
J seruice prouider lihrary, provides a directory-service-specific implemen tation of al l SPI operations and translates resource infor mation back and forth between the I DS entry and the service-provider-specific format. A service provider library must be implemented for each directory service to be supported by I DS. Any directory service or infor mation repository system that Gln provide the I DS SPI semantics may be an IDS service provider.
SPI Semantics
The IDS SPI defi nes the tc>l lowi ng main operations: create, read, search, modi�', discover, and delete. All SPI operations spec i t)' the name of the
IDS com munity upon which to operate. Each
I DS communi ty mai ntains a list of contexts rh:�t specify in which service provider IDS resources of a particular type arc stored and in what location within the service provid er. The SPI uses this comm unity name to retrieve rhe context i n formation that di rects the oper ation to the correct service provider library. \Vith the exception of the del ete operation, which requires an explicitly set context ( to be sure that an expl icitly located object is selected for de letio n ) , i f the cal ler does not set the commun i ty name, the local commu n ity is assumed .
The create, de lete, mod ify, and read functions all operate on a single I DS resource ar a rime. Each, therefore, provides an I DS entry object to idcnr it)' and/or describe the resource.
The create operarjon creates a new l
DS resource i n the d irectory. The create operation specifics the type of
IDS resource to be created, the resource's name, and the I DS attri butes and val ues associ:�ted with the resource. On a successfld create oper:�rion, rhc service provider constructs a unique directory-specific 1wme fi>r the new I DS resource and stores this name in the object's I DS entry. The service provider su bsequen tly may use this name to find the object more qu icklv rather than constructi ng it fi-om the name, resource type, and context information contained in rhc IDS entry.
54
Digital Tcchniul journal
Vol .
H
No.
I 1 996
Before constructi ng the resource in the directory, rhe operation valicbres the I DS entry against rhe schema to ensure that it does not violate the schema .
For ex:�mple, attem pti ng to create a resource without a required attribute va lue pair viol ates the schema and is flagged as an error. Conversely, the delete operation removes the I DS resource from the directory.
The mod i fY operation updates the attri bute :�nd values associated with the resource in the di rectory.
The modi�' operation su pports the following u pdate directives:
•
•
•
•
•
Add a new attribute and value.
Add a new value to :�n existing attri bute.
Replace a val ue of :�n existing attri bute.
Delete an :�ttributc and its associated values.
Delete a value frorn an existing attribute.
Each modi f)r directive is veri ried against the schema before being applied to the directory.
A read operation retrieves a uniquely specified
I DS resource fi·orn the di rectory, translates i t into
IDS entry format, and returns the I DS entry to the call er. The read fu nction is typicaUy used to com pare the di rectory format of an I DS resource to one maintai ned in memory by an application, or to process
IDS resources returned tl·om a search operation one at a rime.
The search fi.I JKtion identities and retu rns IDS resou rces that match the characteristics specified by rhe cal ler. To bound the scope of the search, the ca ller specifies the fol lowing search characteristics: resource type, commu nity n:�me or names to be searc hed, and a fil ter containing attri butes and associated values or value ranges.
The discover operation is called by the I DS system object ro fi nd all comm unities known to a given ser vice provider. Service providers for directory services that support a server solicitation and advertisement network protocol i mplement a discover function. In these directories, servers advertise their presence in response ro network sol icitation requests. The dis cover method uses the d irectory's native solicitation and advertisement protocol to discover local directory servers and then issues the appropriate operati ons to the server to determine if it has defined any I DS com munities. Service providers that do not have a solicita tion and advertisement protocol can i mplement an alternative discovery mechanism such as retrieving the community i n formation from a file or provide no dis covery mechanism.
Construction o f t h e System : Di rectory, Session, and
IDS Entry Objects
The SPI is constructed of three framework objects: the di rectory object, the session object, and the di rectory operation methods of the
I DS entrv object. The directory object is responsible
for service provider initi�1lization and termination,
mai nte nance
of session objects, and com munity dis covery. Each service provider exports one d irectory object to the I DS framework. The session object
implements
all the directorv operations on a service prm·ider. Session objects are obtained fl·om the service provider by means of the di rectory object. The fDS entry directorv operation methods determine the con text if i t
has
not been set, obtain a session object from the proper dirccrory, and d ispatch the operation to the
Jssociated service provider through the session object.
For eHicie ncy, session objects may be cached by the service providers.
Implementation Considerations
Once we bad establ ished our basic approach , we turned our attention to i m plementation decisions.
Client versus Server
Our first consider:�tion was whether to implement this technologv as sofTware executing on a server system or as soft\V�l i"C execu ting on a client system . The server solution had a n u m ber of attractive qualities: it woul d
110t be necess:Jn' to have a l l the native d i rectory cl ient s o n :�II the desktops, Jnd potentially complex pro cessi ng wou ld occur on
:� n :�ppropriate platform .
However, we identified two problems with the server solution . The tirst concerned security. To access rhc di rectory service on bei1Jlf of a particubr user, we would have to imperson:�te that client user on the server machine .
Although this c:1n be done without ex posing securi ty holes, doing so adds :�nother layer of complexity ro the problem. The second problem with rhc server solution w3s thJ.t it requ ired the customer to find a machine tc>r :1nd depl ov
:1 server prior to get r i n g started with the syste m . One of the design goals
WJS to remove lnrriers to directory deplovmcnt, and
II'C were concerned that a server solution would add a barrier. \V<.: s�l\\'
:1 need tor both client- and sen·er based solutions, and since the client solution was sim pler to implement, we chose to stJrt there.
Security
The I DS intcrbccs leav<.: security to the underlying directory services; we did not attempt to abstract
:1 general-pu rpose, �Kccss control or authcntic3tion interbce. The primary reason for this was a conviction that the
V
< l
S t m:�jority of current directory information is world 1-c:ld, and therefore a complex access control i n terbce wJs nor necess:�ry. An access control :�nd
Juthen ticarion l:!yer th:�t was di rectory-service independent would have :�d ded significantly to the complexity of the project, and we chose to postpone this prob l e m . ] D S docs pass requests d irectly to the native direnorv-sen·icc client; IDS does not alter or im personate the user's identity. In that sense , it
pertecrly p reserves the security inheren t i n the u nder lying directory services.
Filter Implementation
The imp lementation of the I DS attribute filter is based on the string fi l ter as defined i n RFC 1 777 -'' The
Lightweight Directory Access Protocol ( LDAP ) string til ter provided :1 convenient internal representation , and we would be J.ble to reuse the LDAP parsing a nd processing code that we had developed as part of an e:�rlier prod uct. We considered using SQL to construct
I DS attri bute filters, but chose not to do this for imple mentation convenience.
Service Provider Considerations
Initially, we though t that developi n g a directory service -independent interface wou ld not be d i fficult.
Most of the required oper:�tions such as read and write are straigh ttorward :�nd obvious. The implementation of such an i nterface, however, proved to be d i fficu lt because the underlying directory services have, in some cases, verv different native capabilities and sen1:1 ntics.
We chose to implement service provider libraries t(x the t(JIIowing th ree types of service providers :
•
•
•
Open D:�tab:�se Connect ( ODRC)-co m pl i:�nt dat:�b:�se
X . 500-based d irectory using the LDAP
DCE CDS
These service p roviders are representative of the types of directories th:Jt exist today. Table
1 high l igh ts some of the d i fferences :�mong the three di rectories. As this table ill ustrates, not all d irectories can nJtivcly su pport the semantics described by the I DS S P I .
I n these situations, we have fol lowed three J!terna tives: ( l ) the service provider l ibr:�ry implements the fu nctionali ty, ( 2 ) the IDS framework implements the fi.1 nctiona litv, or
( 3 ) in a small number of cases, the service provider c:�nnot implement the fu nctionality and remJins less functional .
Some operations c:�nnor be supported natively by only one or J sma l l handful of directory services. For these operations, we requ i re the service provider devel opers to i mplement (or emu late as best they can ) rhe fu nctionality in the specific service provider library for that directory.
For fu nctions that a number of service providers cannot su pport or that arc suHiciently difti cult to i mplement, we provide a common implementa tion or emulation in the IDS framework that service provider li braries c:�n cal l . For example, CDS does not natively su pport an attribute-based search mechanism .
Rather than :�ttempt to implement
:1
CDS search capa bili!J', we chose to provide an I DS ti·:�mework "prune" fu nction thJt :�pplies an I DS filter to :1 list of I DS entries
�md returns onlv those entries that s:�tish' all cond itions of the filte r. Service providers such as CDS can then
Vol . 8 1'\o. 1
1 996
55
56
Ta ble
1
Differences among the ODBC, X.SOO, a n d CDS Di rectories
Functional ity
Distributed di rectory service
H ierarchica l organ izati on of d i rectory in formation
Attrib ute-based sea rch
Attribute val ue-based search
Native schema su pport
User can extend IDS schema
Tra nsactional sema ntics
Tolerant of interm ittent conn ectivity
Provides security mecha nism on connections
ODBC
No
No
Yes
Yes
Yes
No
Yes
No
No x. soo
Yes
Yes
Yes
Yes
Yes
Yes
No
Yes
Yes emul ate the I DS search function by enumerating all resources of a particu lar type and then cal l rhe prune function to pare down the l ist of resources.
The IDS schema implementation is another exJmple of a common capability we have provided tor all service providers to use. Not all service providers support object, schema and, of those that do, fe wer still can sup port user extension of the schema. We chose to allow user extensi bility and i mplemented a service- provider independent schema interface and mechanism.
In a tew instances, we determined that ir wou ld be too expensive i n terms of implementation time ro pro vide a service-provider-specific or a n 1 DS-ti-::llnework implementation of an SPI -mandated function . I n these cases, w e al lowed the service provider to remain noncompliant. For example, a cal l to i nitiate a session to a service provider specifies user name ;�nd p�1ssword argu ments. For those di rectories that su pport user name and password security mechanisms, we preserve that functionality. For directories such as the O D B C service provider that do n o t support these securitv mechan.isms, however, we provide no add i tional sec u rity measures. T h e cost to i mplement a n d deploy such a securi ty mechanism outweighs the gain of h::wing the add i tional features.
In addition, we fou n d that nor all d irectories pro vide the same semantics tor a particular opcr::1tion . for example, when updating a resource, service prm·idcrs handle existence checking of resource attributes d i fkr ently. If req uested to add an attribute value to an attri bute that does not yet exist, one service provider returns an error, while another implicitly creates the attri bute. We worked around problems of this type by carefu l ly specifying the semantics and error cond itions of al l SPI operatiom. Service providers that do not natively support these srr seman tics musr implement whatever add i tional fu nctional ity is req uired ro do so. For exam ple, the CDS service provider req uired additional fu nctions that determined and flagged whether or not a p::1rricular attribute existe d .
I n addition to all errors that are specific to sen·ice providers, '''e retu rn an error rhat is independent of any IDS ti·amework service pro1·id er. This adds another level of consistency across our sen·,ce provider impl ementations.
Current Applications
As with any foundation technology, the proof of its viabilitv l ies with the applications that employ it. In the
PATHWO RKS prod uct, we currentlv han� three appli cations th::tt usc I DS:
•
•
•
Network Connect
I DS Administration
Resource S)'nch ron izer
The Network Con nect applicJtion fi nds and con nects users' printers and tile shares . I t prm·i des a user i nte rface that a l l ows users to browse or search tor tile shares or pri ntns. Through Network Con nect, users can refer to resources by their logical name or their attributes. A single physical printer, with queues on several machi nes or several NOS systems, is presented to users as a single prin ter. Network Connect uses the
I DS API to access the IDS search capabi lities and to translate a pri mer or file sh;lrc's I DS name to its network-specific IUllle to con nect to the resource.
Network Connect mav be accessed through the
Windows version 3.1 Print M<ulJgcr Jnd rile Manager util ities and th rough the PATHWO RKS Network
Con nect utility.
The IDS Administration utility ( I DS Ad m in
) allows a network administrator to m�111;1ge I DS resources and com m u n ities. I DS Admin is int egrated into the
Digi tal ManagcWORKS Workgrou p Administrator for \.Yindows software producr.
1 · ' Ad min creates, mod ifies, �md deletes resources and communities. It a lso al lows users to browse IDS resources and commu nities in the ManageWORKS hierarchy and to search for I DS resources.
CDS
Yes
Yes
No
No
No
No
No
Yes
Yes
Diftiral Technic3l jou rn.\1
Vol. S
No. I 1 996
An ad mi ni strator c111 manage IDS resources manu a l l v th rough the ManageWORKS user interbce or can rely on information p rovided th rough the semi<lllto matic r esource collection utilities called the Resource
Gatherer an d Resource Synchronizer. The Resource
Gatherer periodically collects information about network LAN
M an a g er :md NctWare p ri nters and rile shares. The Resource Svnchronizer utilit\' processes the ga thered i n torn1 a t io n , updating the d ir ec ton
'
.
I t
�1 lso eliminates d u p lic.Hc entries a n d discards informa tion the administrator wishes to ignore. The ga therer and synchronizer allow the directory to be kept up-to d� ne , even if resources arc added or removed through the native NOS in terfaces.
Future Work
In the fu wre, we plan to improve the JDS extensibili t v mechanisms. Cu rrently, a lo c a l cop�· of the s c h ema ex ists on every client. Propaga ting the changes to each client wil l become a problem as users a nd applications extend the sc hema. \Vc arc c o nsidering storing either the schema or a pointer ro the schema in the di rectory.
The current I DS i m plementation run s on both the Windows v e r si on
3 . 1 a nd version
3 . 1 1 operating
S\'Stcms. We are currentlv porti ng i t to Wind o\\'s 95
�111<.l in vestigati ng ports to other operati ng svstcms, such as U N I X .
The implementation docs nor su pport the e n tire
I DS d esig n
:
Although resource c o n t e x t lists arc i mple mented , there is no rcason:1bk way tor a user or adminis trator to crc::nc them. The user intcrf.1cc work ti:lr these katu res in the I DS Ad min application has not vct been com pleted .
Summary
I DS pro vides a mcch:mism tor integrating m u l tiple di recto r y services into �1 s i ngle system . It is predicated
011 the abi l i t y ro define
:.1 common et of directory opc r
::�tions and on rhe type tion of info r m a tion . The implcmenr::� t h r ee verv difkrent servi c e providers - C DS,
X . 5 00, and ODBC - i nci ic::�tes that w e succeeded i n dctining t h e dirccton· o per:-�tions
.
The usc ofiDS i n the
PAT HWO RKS prod uct shows that it does add ress the practical aspects of the problem of integrating mu ltipk directories into a single, logical directory service.
Acknowledgments
We II'Ould like to rhan k rhe manv p a st and present lllcm bcrs of the I DS tc1m ll'ho contributed to the design and i mplementation of the product. Special thanks to K on st
� lnt i no s Baryiames, Anthony
Hin x man, David Magid, Tracy Teng, and Ta mar
\Vexler. vVe wou ld also like ro thank the mem bers of the Directo n
·
Task F orce, Dah Ming C h i u , Dennis
Giokas, and William Nichols.
References l . CC!Tr
Necr! li!lilel ldation
X 50 1
(
1 992 ) and
!nj'omwtion Technology - Ope11 S)'stenls lnter
WIII' Iection-Fhe DirectOl l': ,1,;Jodels. !SO/I EC
9594 · 2 : 1 992
( Gcnn-;1: Intern�rio n�1l Orga ni zation tor Sr;�ntbrdization/l n ternational Ekcrrotec hn ical
Commission, 1 992 )
.
2.
" Naming Concepts" i n Using
Net \!Vare Serl'ices for
NI.Ms ( Provo, Ut;l h : Nove ll, Inc., 1 99 3 ) .
3 . AE'S!Distrihutecl Comput inJ!,- Oirector)'
Serl'ices
( C1111 b1·idge, lV!ass . : Open Softw;ll·e fou ndation,
1 99 3 ) .
4. "S tn:erTalk N;lming Ser� icc" in E.\S Adlllillistrotor �\·
Plan11i11g Cu ide ( Westborough, lvbss. : Bam·an
Systems, fnc., 1 992 ) .
5 . " M icrosoft
Oi n.:crory Services Str�H<.:g)'," � \\' h i tc p•lper from the l� usi ness Systems Tcch nology Scries
( Redmond, W<lsh . : Microsoft Corpor� tion, 1 99 5 ) .
6 . )<J0j)(?l!
Crl l:'
.vJrxi/i'cation. Federated Atlllling· The xr-:v Specijlwtion ( Read i ng, U . K. : X/Open Com pan\' Ltd ., 1 995 ) .
7 .
"
B i n d e n
·
Scn·iccs" i n
Net Ware S)'stelll lnterjctce:
Techuical
Orerl 'ieu·
( Provo, Uta h : Nove l l , l nc.,
1 99 0 ) .
8 . S. R
J d ic a ti
,
" I m ple menting the DIT" i n
X. 500 /) ir ec
IOJy Sen 'ic<'S: Technulop,y and f)ep!o vment ( New
York : Van Nosn·;md Rei nhold, 1994 ) .
9.
P.
Mockapctris, " Dom<Jin NJmcs- Conu.:prs
�nd
Facilities," l nrcrnet Engi n eeri ng T1sk Fmcc, R.l'C:
1034 ( Ncl\'cmbcr 1 9 8 7 ) .
1 0 .
P.
M oc b p ctr i s, " Domain Names- 1 mplcmcntation and Spccitication," l mernet Engineering "L1sk force,
RfC 1 0
3 5 ( Novcmber 1 98 7 ) .
1 1 . Al-.'5/JJistrihuted
Computinp,- Nenwte l'mcedure
Call. AppenchY
A
( C:�mbridgc, M:�ss.: Of)t:n SoFtware
Fou n dation, 1 993 ) .
1 2 . CC!7T f?eco/11 lllc>I1 Ciat ion 208 ( 1992 ) ;l nd t ion Techuolo,L!..)'-Upen lnj(mna
S)'stems lnh·rconl lect ioii
Abstmct S)'ntox Notation One (/iS;V. ! J JSO/I EC:
8824- 2 : 1 992 ( Ccn e\·<1 : 1 n tern<ltional Org:�nizarion t(x
Stantilrdiz;Hi on/l nrcnutional Elcctrorcch nic1l Com m issi on,
1 992 ).
1 3 .
W. Yco ng, T. H m,·cs, ;lnd S. HardclStlc - Ki l l e,
" X . 5 0 0
Lightweight Di rccton' Access Protocol ," Internet
Engincc:ri ng T.1sk Fon.:e,
RFC 1 777
( M�11·ch 1 995 )
.
1 4 . D.
Giobs
� nd ) . Rok icki, "The Design of ,\r!Jn �lge
WORKS: A l ' scr I nred>1cc: hamcwork," f)<� ita/ 'f('ch-
11 ica/fou mal. l'ol. 6, no. 4 (
F:\11 1 994 ): 63-74.
Di!o\iral Tcd1 11ical jou rnal Vol. 1\ :\o l 1996 5 7
Biographies
M argaret Olson lv!ar-garet Olson is a consu l ting software e ngineer in the
Network Software Group. She was the project and tec h nical leJder ror th e IDS development project. for the l ast six years, she has had tec hnical leadership roles in Digital's
Di n:ctorv Services Group. Before joining Digi tal in
1 989, she worked in the networking and distributed. compu ting areas at Apollo Computer. She received a B.A. ( Sigma X i ) h·om Wel lesley College i n network licensing in
1 98 1 .
1 98 8 .
S h e published a paper on
Colin
Strutt
Colin Strutt is a consu l ring software engi neer and tech n ical director ror Teaming Software in rhe Network Software
Group, where he is helping to defi ne new PC- based soft ware prod ucts. Previously, he has held technical leadership rol es i n di rectories, ncrwork management, and term i nal server development, and before rhat l ed product d evelop menrs in F.tllerner servers and DECnet. He joined Digital in
1 9 80 fi-om British Ai rways in the U .K. He received a
B .A. ( honours) in
! 972 and a Ph.D. in
1978, both in com puter science fi·om the U niversity of Essex, U .K. He is a member of BCS and ACM. H e has two parents issued and several patents pending and has pu blished extensively, par ticularly on ma nagement technolog y
.
Laura
E.
Holly
I
.aura Holly is a pri ncipal engineer with the Network
SoFtw<lre Group. She was a key technical contri buro r ro the
I I)S development ef lort. Laura has previously con tributed to the areas of DCE, distri buted system, and knowledge- based system deve l opment. She joined Digital in
1 91\S
,,Iter receivi ng an A . B . ( high honors ) from Smith
College. Lama holds a patent and has publ ished se,·eral papers in the area of k n owledg e
based systems.
58 Dig;ir01l
Technical
Journal Vol .
8
, o.
I 1996
دانلود
تبلیغات
مرتبط