Re: NAT in the original Internet (RE: [wg-b] RE: [wg-c] IAB Technical Comment on the Unique)

[Dave Crocker <dcrocker@brandenburg.com>]

Roeland,

At 03:16 PM 12/19/1999 , Harald Tveit Alvestrand wrote:
>At 10:16 19.12.99 -0800, Roeland M.J. Meyer wrote:
>>Taste issues aside, NAT was part of the
>>original design concept behind the Internet, where local networks were only
>>assigned a few public IP addresses and the remainder were internal-only

This is quite simply incorrect.  The Internet model was always for 
end-to-end/global addressing.  There have been many arguments about this 
issue, over the last two decades.  The outcome has always been in favor of 
providing uniform, global addressing.

NATs are simply the result of failing to increase the address space quickly 
enough.

The issue with LAN addressing is that the extreme popularity of LANs was 
not anticipated.  It was assumed that there would be a relatively small 
number of very large, public, networks, with indvidual hosts attached to 
them.  Emergence of LANS motivated the creation of "classful" addresses, 
permitting a few very large nets, a significant number of relatively large 
nets, and a large number of small ones.  It turns out that what is needed 
is many more small networks than anticipated.

>>addresses. Public IP addresses were only to be assigned to gateway and
>>front-porch systems. NAT is only the enablement of this, already extant,
>>concept.
>
>Hmm.....I'd like a reference (and date) for which "original design concept 
>behind the Internet" NATs were a part of.

Yes it will indeed be interesting to see it produced.

>The Internet grew from a number of ideas, including the famous "catenet", 
>but I think that at the time of the Great Changeover (ARPANet turned off 
>NCP and turned on TCP/IP, all in one night - Jan 1, 1983), the idea that 
>all end-nodes had globally unique addresses was a fairly well established 
>part of Internet "orthodoxy".

The concept of the catenet was to link together existing networks, such as 
the Arpanet and a number of its siblings in Europe.

Perhaps the confusion is about this intent to run IP on top of such 
networks -- think of them as equivalent to X.25 -- without touching the 
lower-layer networks' addressing.

Over time, we are increasingly seeing the disappearance of the lower 
layered nets, with IP being both the network sub-layer and the internetwork 
sub-layer.

d/

=-=-=-=-=
Dave Crocker  <dcrocker@brandenburg.com>
Brandenburg Consulting  <www.brandenburg.com>
Tel: +1.408.246.8253,  Fax: +1.408.273.6464
675 Spruce Drive,  Sunnyvale, CA 94086 USA