Re: Correction of my last two comments "the net"

[David Farber <dave () farber net>]

Begin forwarded message:

From: Dave CROCKER <dcrocker () bbiw net>
Date: September 13, 2009 9:07:39 AM EDT
To: Dan Lynch <dan () lynch com>
Cc: "David P. Reed" <dpreed () reed com>, Steve Crocker  
<steve () shinkuro com>, Richard Bennett <richard () bennett com>, Dave  
Farber <dave () farber net>, ip <ip () v2 listbox com>, Gordon Peterson <gep2 () terabites com 
>, John Shoch <shoch () alloyventures com>, Harold Burstyn <burstynh () iname com 
>, Lauren Weinstein <lauren () vortex com>, Paul Robichaux <paul () robichaux net 
>
Subject: Re: Correction of my last two comments [IP] "the net"



Dan Lynch wrote:
> David, I like your take on the cutover date --  a fading away of the  
> Arpanet as it was loved and used.  The “Internet” had been running  
> for years by then as you have previously described.
...
> On 9/12/09 4:38 PM, "David P. Reed" <dpreed () reed com> wrote:
...
>    The idea that 1983 is the "birth of the Internet" rather than the
>    fading away of the ARPANET seems to be somewhat important to people
>    who want a date.  IMHO, 1974 was the birth of the Internet we have
>    today, because that is when Cerf and Kahn defined the framework  
> for it.

Jan 83 is significant because it marks the moment the Internet  
officially moved from experiment to production service.

Most inventions fail, so it's worth noting when a critical success  
milestone is achieved.  I don't have a good suggestion for  
terminological distinctions for the markers, but think that the  
differences between Internet invention (1974), testing (1976) and  
production (1983) are essential.

Having not participated in any of these milestones, I could easily  
believe that the production one actually occurred earlier.  As noted,  
1983 marked a removal of Arpanet Host-to-Host protocol service, not  
the start of TCP/IP as a production service.




Steve Crocker wrote:
...
> Frank Heart was vehemently opposed to us including a checksum at the
> host level.  He thought it was an unnecessary inefficiency and would
> make his network look bad.  We backed down, a decision we came to  
regret
> soon enough.
...
> If I had been older or smarter or something, we would have included
> checksums, retransmission and reordering in the Host-Host protocol,  
the
> 1822 specification notwithstanding.


Probably not.  I don't think that folks knew enough yet to win that  
argument or had enough sway to force it.

What you are describing was a very basic paradigm debate. Deciding how  
much to put into the underlying infrastructure remains a basic point  
of dissension with really bright people continuing to be on both sides.

The debate is pursued either as a divide-and-conquer parallel effort,  
with one side eventually gaining more operational strength, or  
occasionally having both continuing forever; or it is pursued with one  
side having stronger political force and managing to kill its  
competition.

The lessons that showed the massive superiority of putting serious  
reliability mechanisms into the hosts rather than the net took at  
least another 10 years to gain much sway  Note that, for example, the  
rather bright folks at PARC didn't get it for XNS.

While I was at Ungermann-Bass, in 87, we ran into the debate in a  
painful way. Its core products ran an XNS derivative and I was  
managing early development of a TCP/IP alternate suite.  There was a  
major fire-drill caused by a customer discovering that it was losing  
critical accounting data that was shipped across its UB net.  This  
turned out to be an interface failure and was incredibly difficult to  
debug.  After it was resolved, those of us on the TCP project noted  
that with XNS it caused major damage whereas with TCP it would merely  
have tripped a retransmission counter, but otherwise have had no  
visible impact on service.

And for reference, the debate continues in the IETF today.  As another  
devotee of "put as little into the infrastructure as you can get away  
with", I of course view the "keep the end points as simple as  
possible" folks as impossibly naive and misguided...

d/
--

 Dave Crocker
 Brandenburg InternetWorking
 bbiw.net




-------------------------------------------
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