Re: Parttime Gateway via Dialup SLIP... ...HELP!

[mikebw@bilow.bilow.uu.ids.net (Mike Bilow)]

I am directing this reply to a reply as a guess; I received a private CC of the
reply and I don't remember which mailing list it was from.

On 94 Dec 26 at 17:15, Michael Stenner <mstenner@haven.uniserve.com> wrote:

MB>>Since 44.x.x.x is a Class-A address, the Internet considers 
MB>>it to be a net whether you think it is or not.  If the net is 
MB>>formed by wormholes or some other means, then it still looks 
MB>>like a net.  The fact that it is not a fully
MB>>connected net does not mean that it is not a net.  Rather, 
MB>>it is a net that does not work perfectly because it is still 
MB>>under construction.

 MS> 1.      Please don't confuse DOMAIN with NETWORK!

I didn't, but I think you did.  :-)

 MS> a.     I agree 44.x.x.x is a Class-A address, but 
 MS> this does not imply a
 MS> network - it's considered a DOMAIN!  Whether there is one 
 MS> computer or an entire network under a DOMAIN is irrelevent.  
 MS> The same is true of configuration...   ...and TCP/IP could 
 MS> care less!

The point is that the connected Internet wants to route all frames for a
Class-A network to a single router, in this case MIRRORSHADES.UCSD.EDU.  What
happens after the frames get to that router is "behind the curtain" from the
point of view of the connected Internet.  In other words, the rest of the
Internet assumes that the Class-A network is actually a fully connected network
with a single point of contact.

 MS> (The following is not patronization, but education)

OK.

 MS> b.     The proof?  Telnet to rs.internic.net (198.41.0.5 -
 MS> Internet's Central
 MS> Repository) and try "whois 44".  Note the "NET-AMPRNET" and
 MS> "AMPRNET".  The first DOES identify this DOMAIN as a 
 MS> NETWORK; the second identifies the DOMAIN.  Now try "whois 
 MS> AMPRNET".  Your attention is drawn to the "AMPRNET-DOM".  
 MS> This record links the DOMAIN NAME to the 44.x.x.x DOMAIN 
 MS> (#.0.0.0 - is the highest level within a DOMAIN), and 
 MS> identifies the DOMAIN SERVERS.  (If one looks familiar, it 
 MS> should.)  Now try "whois ampr.org".  This record links the
 MS> mnemonic suffix "ampr.org" to the 44.x.x.x DOMAIN.

I must respectfully disagree with you about this.  NOTHING at Internic links
"AMPR.ORG" with any IP address, whether in the 44.x.x.x Class-A net or not.  In
fact, Net 44 is slightly unusual in that its authoritative name servers --
UCSD.EDU and TROUT.NOSC.MIL -- are not actually in Net 44.  Only resource
records maintained by these authoritative name servers make the link between
AMPR.ORG style names and IP addresses.

I suppose you could argue that a block of IP addresses is a "domain" in the
sense that someone has traceable administrative authority for them, but that is
not the common use of the term in TCP/IP.  A network is generally considered to
be mappable into a pseudo-domain (IN-ADDR.ARPA) for purposes of back resolution
using PTR resource records, but that is not the same thing as saying that a
network is a domain.

It is merely administrative policy that requires AMPR.ORG domains to be mapped
to Net 44 addresses and vice versa.  The administrator is free to map whatever
he wants to wherever he wants.  I really am at a loss to follow your chain of
reasoning regarding "whois" queries.

 MS> c.      My arguement remains valid.  Furthermore...
 MS>         
 MS> d.      The term "#net" is a holdover from the ARPAnet 
 MS> days, nicknames for
 MS> it's sub-networks (rarely used anymore).  This is 
 MS> because ARPAnet didn't
 MS> impliment a DOMAIN NAME convention.  To my knowledge, 
 MS> ARPAnet never carried encapsulated AMPRnet traffic.  By 
 MS> continuing to use the term "44net"
 MS> you break with current naming conventions, create 
 MS> confussion, and could potentially find some old 
 MS> military-types taking offence to this infringment.

You cannot split up or delegate network routing.  Sure, you can delegate back
resolution with NS resource records for pseudo-domains x.x.n.44.IN-ADDR.ARPA,
such that name service is delegated to different machines for different values
of "n," but that is still not too important.  Net 44 looks like a fully
connected network to the Internet -- and I do mean "network" -- because all IP
frames addressed to it must be routed to a single point.  Techniques such as
AXIP (RFC1226) have been used to connect disparate pieces of Net 44 using
wormholes through the rest of the Internet, but that only supports the illusion
that Net 44 is a fully connected network.

I'm talking about where IP frames go, not what gets typed into name servers.
 
-- Mike