[ih] Resource sharing (was: Re: First file transfer on ARPANET)

[John Klensin]

On Fri, Dec 21, 2012 at 4:51 PM, Miles Fidelman
<mfidelman at meetinghouse.net> wrote:
>>
> Maybe I'm confused, but I'd consider all of these to be examples of  distributed computing:
> - distributed databases - both SQL and noSQL (can you say map-reduce?)
> - federated geodata systems
> - some of the "big science" grid systems (e.g, the Large Hadron Collider Computing Grid)
> - mashups (both client and server side)
> - the Erlang/OTP run-time environment

And Dave Crocker wrote about cycles in distributed computing work.

Most of these are more or less "traditional" client server
applications, just as much of the web is remote login in disguise,
albeit aided by downloadable software to the local devices (more than
the IBM 3270, or even the "intelligent terminal" ideas of the 60s and
70s, but maybe not a big conceptual leap).  IMO, the shared-resource
activities that John Day and I are describing were somewhat different,
involving having the local software resolve user interface issues,
determine what resources were out there that could best meet the
requirements of the particular user and application, and then seeking
those resources out, using them, and making them look local.to a
non-computer-expert user.  It was an explicit non-requirement that all
of the systems and hosts involved be homogeneous with regard to
operating systems, languages, data structures, or accessing software
-- although commonalities obviously made things easier, the assumption
was that the most important "remote" resources would be running
systems highly tailored to their needs.   All of that, as John points
out, pretty much requires an application-oriented operating system (or
operating system layer) on the host(s) directly supporting the users
--exactly a key component of what we were trying to build, although
our primary focus was closer to the Multics-like "computer utility"
(with ability to accommodate remote resource-sharing) than to "network
operating system". (If anyone is curious, the most accessible
reference on what we were doing is probably Dawson, Ree, John C.
Klensin, and Douwe B. Yntema, “The Consistent System”, The American
Statistician, 35, 3 (August 1980), pp. 169 176.)

As far as I know, very little has happened with that area of work
since our efforts, those that John describes in Illinois, and a few
smaller-scale ones of the same period, fizzled out.  Perhaps they have
become irrelevant.  For example, those efforts certainly never
anticipated what would have been considered more-than-mainframe
horsepower on handheld devices or multiples of a terabyte of storage
on a desktop machine.  There is also far less diversity in operating
systems, and probably in data formats, than we were dealing with at
the time and that might also further reduce the requirement.    My own
guess is that, as with operating systems and maybe even with
application-oriented programming languages have still not caught up
with where we were, or were headed, in the 70s.  I also believe that
availability of levels of processor power, memory, and network
bandwidth that were almost inconceivable 40 years ago has changed how
we think about quality and the costs of buggy and inefficient systems,
not necessarily for the better.

But I think we do ourselves and the community a disservice by
confusing those resource-sharing ideas with remote login, remote
access, or even other flavors of client-server computing and even
largely homogeneous distributed computing efforts (I'd add SETA and
similar network-based collaborative and distributed efforts to Dave's
and Mile's lists) and thereby losing sight of the former.

best,
    john