NetNews Usenet Archive 1993 #1/NN_1993_1.iso/spool/comp/database/theory/673

Text File | 1993-01-07 | 3.2 KB | 70 lines

Newsgroups: comp.databases.theory
Path: sparky!uunet!stanford.edu!kronos.arc.nasa.gov!butch!iscnvx!news
From: [e-mail]
Subject: New Model
Message-ID: <[e-mail]>
Sender: [e-mail] (News)
Reply-To: [e-mail]
Organization: LMSC, Sunnyvale, California
Date: Thu, 7 Jan 93 18:31:27 GMT
Lines: 58

I read the latest trade magazines on the subject of new distributed
processing technology and ATM.  They are close, but haven't quite gotten
it as far as it should be.

The current client server applications work just fine with the current
networking systems.  The issue here is replacing the client server model
with the semantic data model, and whether the replacement can be
accomplished without some virtual path network technology.

Example:

  Consider putting the entire corporate literature database under the
semantic model with schema such as:

   Reports have text, abstracts, key words, authors, they reference other
reports, and are attached to product development teams.

   Authors have resumes, and have worked on other product development teams,
and have written other reports.

  Given this schema, it can be construed as a network connection of data,
and queries are construed as traversals and extractions from the data.
For example: Which reports reference other reports which use a specified
key word.

  Data can be entered into the corporate semantic model regardless of its
location, classes can be indexed across the corporate semantic model, and
class instances can be discovered by some exploration of the corporate
network.

  When the semantic model has certain properties, like: Single valued
traversals, existence of property values, and type consistency
along traversal; then a set of equivalence and inference axioms result,
and these axioms allow the definition of a traversal protocol of which
Wave is the most obvious example.

  The axioms define associative, transitive properties of paths, and
their related traversal programs.  Included is a kind of distributive
property which allows multi-variate recursive queries. The distributive
axiom requires two operations, the path splitting, and node hopping
functions.

  This is a different model then the client server model, and it requires all
the flexibility of a programmable messaging layer, it needs the self routing
capability of an application specific network layer.

  The semantic theory, as far as is published, seems to be rapidly closing
on an access strategy language which has a syntax almost totally defined
by theory, hence differences in implementations will be minimal.

  In order to short circuit the understanding process, I will try to
cross post some ideas in both the comp.database.theory and comp.dcom.cell-relay.
We want the data base theoreticians to include virtual path networks into
their model.  Hopefully they will begin to see that a traditional router
or bridge, is simply an index on some class; that a shortest path is simply
a semantic reduction.  Then they can begin to think about the network
management issue in terms of semantic theory; and conversely, think about
the virtual path networks as a connectionist machine.