home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Danny Amor's Online Library
/
Danny Amor's Online Library - Volume 1.iso
/
bbs
/
society
/
society.lha
/
PUB
/
isoc_news
/
1-1
/
n-1-1-040.31.1b
< prev
next >
Wrap
Text File
|
1995-07-21
|
4KB
|
77 lines
N-1-1-040.31.1 Resource Discovery Beyond X.500, by Michael F.
Schwartz*, <schwartz@latour.cs.colorado.edu>
The Internet connects thousands of sites and millions of users around
the world. As it continues to grow and offer new types of services,
being able to locate and make effective use of the available resources
becomes increasingly difficult. To address this issue, the CCITT has
developed a directory service specification called X.500, as an OSI
application layer standard.
X.500 describes a hierarchical collection of servers, with provisions
for caching and replication. Each participating site maintains
directory information about resources at that site, as well as
administrative information needed for traversing the tree and
maintaining proper distributed operation. Unlike the TCP/IP Domain
Naming System, X.500 supports authenticated runtime updates, and
stores typed data using a structured schema. Field trials conducted
by Performance Systems International and the Field Operational X.500
Project demonstrate that there is significant interest in deploying
X.500 servers at institutions around the world. Tools exist to ease
the task of searching the tree, and to make long-distance operations
more efficient. Graphical client interfaces exist for a number of
platforms.
X.500 is an important standard with growing momentum, but it is not a
complete solution to the resource discovery problem. Its current use
focuses primarily on providing a "white pages" directory of Internet
users. However, over time X.500 will need to accommodate many other
types of resource discovery. Consider two realms very different than
user directories: commercial network services, and wide area
distributed file systems.
In a commercial network service environment (such as airline computer
reservation systems), the resource discovery mechanism should support
fair access among competing information providers. This issue will
heat up significantly in the next few years, as the U.S. Regional Bell
Operating Companies enter the information services market, and the
Internet begins explicitly allowing commercial traffic.
Supporting resource discovery in a distributed file system requires
support for two distinct problems. First, how does a user discover a
needed resource based on an attribute-based description, such as
source code for a particular X-window system application? Second, how
does a user locate an appropriate instance of this resource, from
among the many replicas available (e.g., by anonymous FTP)? This
decision should consider network bandwidth and version information
about the file, and eventually should also consider cost and policy
routing considerations. The popular Archie system developed at McGill
University addresses the first of these problems. The second problem
is currently a topic of research.
Beyond supporting different types of resource discovery, other
problems arise in trying to organize a widely shared, broad
information space. While the hierarchical organization used by X.500
supports scalable decentralized administration, hierarchies become
convoluted as an increasingly wide variety of resources are
registered. Moreover, hierarchical information is only efficiently
searched according to its primary organizational attributes (country
and organization in the case of X.500). Searching for resources
according to other criteria (such as the functionality of a software
package) is inefficient. Inserting cross links between parts of the
tree according to such criteria does not adequately solve this
problem, since the information is still physically distributed in a
fashion that does not permit efficient searching. Moreover, creating
such links requires a large amount of manual administrative effort.
There are a number of research efforts under way to address the
problems discussed here. In time, the ideas introduced by these
projects may find their way into future versions of X.500.
*Assistant Professor, Department of Computer Science, University of
Colorado - Boulder