N-1-3-040.31.2, "Internet Uniform Resource Indentifiers", by Alan
Emtage*, <bajan@bunyip.com>


We in the Internet community are now seeing the cold, hard reality of
what it means to have exponential growth and not one of Sacred Seven
Layers has been spared in this attack.  One of the most obvious
examples of this from the user's point of view, has been the explosion
in Information Services.  From archie to Z39.50, users are being asked
to navigate through hundreds of Gigabytes of data to try and find just
that information for which they are searching.

In moving from an Internet world composed of weenies & wizards in the
darkened offices to John Doe on Maple Street, a fundamental change
occurs: suddenly we will have millions of "librarians" and
"publishers" who won't be (and can't be expected to be) trained in the
knowledge of those professionals.  Recently, much attention has been
focussed in the Information Systems community on how to identify,
locate and access the (potentially) millions of resources on the
network.  Although there remains much work to be done, consensus seems
to have been reached on the broad outlines of how such a system would
work.  The idea in its most basic form is that the location and access
component is separate (and external) from the identification (internal
content) of any object.  Having done this, the following scheme has
been developed.

Uniform Resource Locators (URLs) will specify a well-known format in
which the various information services interoperate and exchange data
about the location and access methods for a specific resource.  Thus
WAIS and Prospero would be able to parse and understand references to
the the same resource (document, service, etc.) regardless of in which
system the information is embedded and the internal content of that
information.

Uniform Resource Serial Numbers (URSNs) on the other hand will be
concerned with the identifying the actual data itself.  There is
currently no way to know if two resources with the same name (e.g.,
documents with the same filename) actually contain the same
information.  Conversely, two documents with the different names may
contain the same information, but in a different format (for example,
ASCII or PostScript).  Ideally we would also like to know how two
objects may be related.  Is one derived from the other?  Is it the
same document with spelling errors corrected?  Is there a way of
uniquely "signing" the object?

Finally, Uniform Resource Identifiers (URIs) are comprised of the
union of URLs and URSNs and would uniquely specify any resource on the
network.  Work is now beginning in the Internet Engineering Task Force
(IETF) to describe and define the problem and to try to work out
potential solutions.  A draft document describing the architecture of
the URL system as discussed at the last IETF meeting is currently
being written by Tim-Berners Lee of CERN (timbl@nxoc01.cern.ch) and is
due out in the next few months.  The URSN is an inherently more
complex problem and it is expected that there will be much wailing an
gnashing of teeth before the final form of this is known.

In many respects we are re-inventing and reworking much of Library
Sciences in the Internet environment, with one significant difference:
scale.  Although "The Library" deals with amounts of information
several orders of magnitude over what is currently available on the
Internet, they effectively have centralized control over how books are
classified and numbered (e.g., the ISBN designation).  On the
Internet, such tight centralized control would soon become a
bottleneck in the system, and in addition, information cannot always
be easily defined in terms of books, films or tapes (a running process
for example).  Perhaps the decentralized allocation methods of the
Domain Name System could be adapted to the purpose.  Again however,
the problem is that while there may be 10 million machines on the
network 3 years from now, there may be 100 million users and 1 billion
objects; a difference of two orders of magnitude.  A tall order for
any system.  This work has the potential to have a tremendous impact
on how we use the Internet in years to come.  Who knows how it will
turn out?  In any case, there'll be a lot of fun in just getting
there.


*VP Research & Development for Bunyip Information Services