re A Ridiculous Failure of Critical Infrastructure

[Dave Farber <dave () farber net>]

Begin forwarded message:

> From: Steve Goldstein <steve.goldstein () cox net>
> Date: December 2, 2009 1:30:57 PM EST
> To: dave () farber net
> Cc: Steve Goldstein <steve.goldstein () cox net>, ip <ip () v2 listbox com>
> Subject: Re: [IP] A Ridiculous Failure of Critical Infrastructure

> I was at NSF-CISE-DNCRI during those times.  I kept a plot of the  
> "size" of the various parts of the Internet, viz. the  
> "commercial" (I think it included .com and .org, and I am not sure  
> about .net) vs. the "non-commercial" (total of .edu, .mil,  
> and .gov).  By just about any measure (number of registrations,  
> traffic flow, size of pipes), the non-commercial component kept on  
> growing almost linearly (my recollection), while the commercial  
> component grew rather slowly in the first few years of the 1990's,  
> crossed the non-commercial component in 1994, and then shot up much  
> like a parabola (dare I say exponentially? I never actually fit an  
> equation to the curve...just eyeballed it) and thoroughly eclipsed  
> the non-commercial within a few years.  I am pretty sure that I did  
> not include non-US domains in those plots.  But, I think that the  
> main point is still valid: 1994 was the "crossover" year, and from  
> then on the Internet had commercialized.  RIPE NCC actually  
> published monthly statistics of registrations, at least during the  
> first half of the 1990's.  They might still have those numbers lying  
> around if anybody wants to look.
>
> Next, NSF did not "give its own backbone to private industry."  What  
> happened was that (1) the needs of the research and academic  
> community that the NSFnet backbone served outgrew NSF's ability to  
> pay for sufficient capacity, and (2) commercial ISPs complained to  
> Congress that the Government (NSF) was unfairly taking away their  
> potential customers by serving the R&A community.  So, the decision  
> was made to go on subsidizing connections to commercial backbones as  
> long as the chosen backbones met up at at least three designated  
> exchange points so that all supported institutions could reach all  
> the others by virtue of traffic exchanges (NAPs, or network access  
> points).  This gave an early boost to the concept of peering, though  
> the NAPs were not the only peering points.
>
> In about the same time frame, NSF solicited bids for a very high  
> speed backbone network *service* (or vBNS), note the emphasis on  
> "service."  This was to be a 45 Mbps backbone service, which could  
> be implemented in such a way that the the purveyor could also sell  
> services to other customers. And, this was because NSF could not  
> afford to pay for its own 45 Mbps backbone.  But, the vBNS was not  
> for normal Internet traffic, but for support of actual research  
> projects requiring more than every day uses such as e-mail and  
> browsing.  The vBNS was furnished by a new company that was set up  
> for that purpose with Al Weiss, a former IBM VP as its CEO, ANS  
> (which I later realized was SNA spelled backwards...but I digress).   
> And, we set up STARTAP, an international high-performance R&A  
> exchange point in Chicago so that the newly formed high-performance  
> networks of other countries could exchange traffic with each other  
> on a bi-lateral (no US AUP) agreement basis.
>
> And, again, in about the same time frame as the vBNS (I am typing  
> this all from memory, and my memory for dates is lousy), the  
> universities were dissatisfied with the service that they were  
> getting from the commercial ISPs (underprovisioning, congestion), so  
> Internet2 was birthed as a private membership organization to  
> acquire a private university backbone.  Internet2 lives on, as does  
> National Lambda Rail (NLR), a very high performance private backbone  
> network that serves a customer base similar to and overlapping with  
> that of Internet2.  Readers will recall the notice that Glenn  
> Ricart, who originally headed up one of the original NSFnet regional  
> networks, SURAnet, is the newly appointed CEO of NLR.
>
> And, other U.S. Governmental research networks are active, e.g..  
> Energy's ESnet and (still alive?) NASA Science Network (NSN).  And,  
> the STARTAP morphed into STARlight, an international lightwave  
> exchange point located on the lakefront campus of Northwestern  
> University in Chicago. STARlight is but one of several international  
> lightwave exchange points used by international R&A networks that  
> comprise the Global Lambda Integrated Facility (GLIF), a cooperative  
> that serves computationally intensive research needs globally.  And,  
> much to my dismay, several of the GLIF member countries (NL, KR, CA,  
> JP, ...) have left the US in the dust when it comes to advanced  
> infrastructure provisioning.  The US has a lot of catching up to do.
>
> Whew, I started out to type a few sentences, and I ran off at the  
> fingers.  Hope that what I wrote is reasonably accurate, and that I  
> have not bored you.
>
> --Steve
>
> > > > Let's be clear about three things:
> > > >
> > > > First, by the time the NSF walked away from the Internet the  
> > > > network had a very decentralized operational structure. Even  
> > > > during the research days, the NSF only managed the backbone,  
> > > > while regional cooperatives managed the regional networks, and  
> > > > thousands of



-------------------------------------------
Archives: https://www.listbox.com/member/archive/247/=now
RSS Feed: https://www.listbox.com/member/archive/rss/247/
Powered by Listbox: http://www.listbox.com