[ih] Dave Farber's DCS

Toerless Eckert tte at cs.fau.de
Wed Jul 22 17:48:44 PDT 2020

Would also be interested if there was some more info on DCS. 
I am sure that basic communications primitives for distributed
apps have been reinvented over an over similarily, but never
tried to figure out their history. Just reminded me of a
recent attempt to standardize them on the basis of discovery,
synchronization and negotiation. 

Bidding based distributed applications have of course been used in the
Internet industry much longer than what you see now. This round is just about
use in "public" services. If you think instead of even 15 year or older
large scale web applications implementations, they too worked on that scheme. E.g:
rendering of a web page would be be done by a collector putting
out bits for all the different parts of that web page, and or each part
there would have been hundreds of nodes with the capability to perform
the job and based on their load one would be selected to actually do it.

And of course all this web stuff from the last 20 years was also based
on distributed system designs from the decades in before. I remember
CORBA as one of the IMHO (in)famous frameworks that i saw as late as
the 2000'th but which i think attempted to wrap up those distributed
communication paradigm designs into an OO architecture, when OO was
still going up the hype curve (1991).

On Wed, Jul 22, 2020 at 04:48:11PM -0700, Karl Auerbach via Internet-history wrote:
> When I was at SDC during the 1970's I met Frank Heinrich (who eventually
> became my Best Man at my wedding.)  Frank was a student of Dave Farber on
> the Distributed Computing System (DCS) at UC Irvine during the late 1960's.
> DCS used distributed services (via a ring network) to build the chunks of an
> operating system.  It used a bid-and-contract system to bind the pieces
> together - if you needed file storage you put out a bid for storage and
> machines that had that facility would send back a response, with a cost.
> Then there was a contracting phase to bind the relationship.
> (IBM had to concede to DCS that IBM did not invent the token ring network
> concept.)
> Frank H. taught Dave Kaufman and me about DCS.  That had a lot of influence
> on some of the things we were working on under wraps of government secrecy.
> Dave K. and I were fascinated with the idea of extending capability-system
> like protection domains, with privilege delegation, across the net to create
> encapsulated DCS-like network services.  (This was before one of our
> consultants, Whit Diffie, did his public-key cryptography stuff.)
> I don't think that DCS had much of a practical impact on the net.  That is,
> not until the last decade or so.
> That concept kinda went dormant for a long time.  But it came back with the
> rise of HTTP/S based web APIs, such as in Amazon's AWS and many of Google's
> APIs to things like graphing tools and such.
> It has also come back with people using Docker containers as services that
> are created and scattered around the net as needed, often with DNS or a
> higher-level naming service used to locate and bind them.
> Is there a description of DCS that is better then the inadequate snippets
> and fragments that one finds occasionally?
> (BTW, I did a multi-hour interview with Dave Farber some years ago and a
> sizeable part was about DCS.)
> 	--karl--
> -- 
> Internet-history mailing list
> Internet-history at elists.isoc.org
> https://elists.isoc.org/mailman/listinfo/internet-history

tte at cs.fau.de

More information about the Internet-history mailing list