[ih] First file transfer on ARPANET

[John Klensin]

On Wed, Dec 12, 2012 at 1:24 PM, Noel Chiappa <jnc at mercury.lcs.mit.edu> wrote:
> You're splitting a rather fine hair (although I concede the accuracy of that
> hair): The TELNET protocol may _be_ a terminal device driver protocol, but it
> was mostly _used_ for remote login - and the server to which one connected at
> port 23 using it was a remote login service...

Noel, to split another rather fine hair, but we've got several NVT
protocols (Telnet without option negotiation and most of them defined
that way) which, in the aggregate, have probably been used more than
Telnet itself, especially in recent years.  Prominent members of the
group include Finger, Whois, and FTP itself but I remember a lot of
nearly ad hoc arrangements that probably never made it into the RFC
series or other official network documents.

The original claim about a focus on remote login is incorrect in
another way.  To the extent to which there was a focus on anything
other than getting the bits moving, it was on resource-sharing (and
resource-sharing by users of the network, not just for development
purposes).  Certainly both remote access to distant machines and file
transfer were part of that picture.  But Lick was pushing us (and I
presume other applications-oriented groups) by 1970-1971 toward
demonstrating "real" resource-shared applications work.  The goal was
that users would sit in front of local devices, dealing with their
familiar local applications interfaces, and carry out calculations and
other actions transparently as to whether programs and/or data were
local or on remote machines.   The applications needed to be able to
bring data to programs,  programs to data, and the presentation of
both to the user.  Conversion of data formats and user interfaces
would, necessarily, be invisible to the user as well and we were
trying to build on some of Selfridge's ideas about agent programs as
user interface tools for the latter.   We did some preliminary
demonstrations of some applications working that way by about 1975 (in
conjunction with some efforts at Illinois -- I'd have to paw through
now-ancient annual reports to reconstruct just when).  Descriptions of
those efforts never made it into the RFC series (probably for multiple
reasons including that it was real applications work, that we had
limited resources for those efforts, and that the work was not
supported through IPTO).

I guess one major thing we learned from those demonstrations was that
the network and most host machines weren't nearly fast enough to make
the ideas realistic for any but the most persistent and patient users.
 I've never understood what finally killed the Datacomputer project,
but have assumed that the speed/performance issues were part of that
too.

   john