[ih] Success has many fathers but failure is an orphan

[Louis Mamakos]

There were some supercomputer/HPC networks running around at the time, USAN ("University Satellite Network" or something) was one I remember, and there was a DoE network, I think, that was DECnet based.  And the DECnet based NASA SPAN network, though Milo was at work with the IP-based NSN, too..   We had a few of these land at the University of Maryland, College Park campus and it was a tricky proposition to figure out how to plumb these things into a campus network infrastructure.  While the operators of these networks might imagine a purpose-built network that was run somewhat in isolation, the burden of integrating these networks at all the "edges" into their customer's networks was still there and very problematic.  

While we still hadn't figured out generalized network connectivity over multiple backbone networks at the time, at least the Internet protocol suite imagined such a thing, and it was the NSFNET (and multiple NSF sponsored regional networks) that forced the architecture to really embrace multiple "core" networks, rather than the implicit ARPANET core ruled by EGP at the time.

Louis Mamakos

On Jan 5, 2013, at 10:27 PM, Vint Cerf <vint at google.com> wrote:

> I have no reason to dispute the facts in the story, either. However, the writer does fail to recognize the pioneering nature of the CSNET (it adopted TCP/IP thanks to Larry Landweber's 1980/81 intervention before Dennis Jennings made the same decision for the NSFNET around 1985. If I am remembering correctly, Dennis was involved with the super computer effort at the time he recommended that the nascent NSFNET also make use of TCP/IP. In Fall 1986, then-Senator Gore held a hearing at which Bob Kahn introduced the term "information infrastructure" and Senator Gore asked whether an optical fiber network should be constructed to link the Supercomputer centers together. The head of CISE at that time was Gordon Bell and he convened a Feb 1987 meeting in San Diego that led to the proposal for the National Research and Education Network program. I also seem to recall that the supercomputer center directors lobbied unsuccessfully to build their own, disconnected networks on the grounds that performance required specialization and control by each center [perhaps someone on the list can clarify that hazy memory]. They were overruled (assuming I am remembering this correctly) in favor of an NSF-wide network.
> 
> vint
> 
> 
> 
> On Sat, Jan 5, 2013 at 9:07 PM, Guy Almes <galmes at tamu.edu> wrote:
> Alex,
>   I was UW Computer Science during the time period he writes of and I was not aware of him.  But, myopia aside, his story rings true with me for several reasons.
>   First, whatever the reason was, the fact that NSF funded and (more or less) organized an effort to connect all the US universities to the (NSFnet) Internet was clearly of huge importance to the growth of the Internet.
>   Second, the specific driver, enabling effective remote access to the five or so centers from the 200 or so research universities, was both a demanding application and one that gave the NSF a defensible reason for funding and organizing the NSFnet effort.
>   Third, that remote supercomputer application presented a clear motivation for much higher end-to-end capacity that the ARPAnet-based Internet that existed in 1984.  In short, moving gigabyte files motivated T1 performance levels.
>   Fourth, this story helps one understand the substantive ways in which Sen. Gore's efforts made a difference.
> 
>         -- Guy
> 
> 
> On 1/5/13 5:38 PM, Alex McKenzie wrote:
> http://www.analogsf.com/2013_03/altview.shtml
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://elists.isoc.org/pipermail/internet-history/attachments/20130105/c198d537/attachment.htm>