The IP over ATM Mailing List Archive by date

[Ajay Joseph <ajayj@panix.com>]

On Mon, 13 Mar 1995 atmpost@panix.com wrote:

> All:
> 
> Reading over the RFC-1577 and NHRP specs over the weekend, a thought
> occurred to me.  Has it been discussed, eliminating the need to translate
> IP address to ATM End-System Address?
> 
> As you probably know, there are currently 4 forms of ATM End-System Address.
> What if we made up a fifth, OSI NSAP containing IP address?
> 
> Probably I'm missing something here, because it seems too simple.  But
> I believe that some early ATM networks will use IP routing internally
> (because it exists, and fancier things don't) and in fact convert
> other forms of addresses into IPv4 addresses internally.  So it seems
> a shame to do an ATMarp, only to convert IP to ATM, then have the network
> immediately reverse this process.
> 
> Even in the future, when fancier routing protocols are available, I see
> no fundamental reason why a network couldn't support IP addresses directly.
> If I have to support 4 address formats, what's one more?
> 
> I assume one drawback is standards support.  Let's hand-wave that one
> for the moment.  Are there technical reasons for *not* doing the above?
> 
> Just as a strawman, here's a suggested mapping of IPv4 and IPv6 (the
> latter stolen from draft-carpenter-ipv6-osi-00.txt) to AES address.
> 
>      +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
>      |A | ISoc| IPv4 addr |////// not used? //////|    ESI    | S|
>      +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
> 
>               |-------------- H0-DSP -------------|
> 
>      A      = Authority and Format Identifier (AFI) = 47 (ISO ICD format)
> 
>      ISoc   = value indicating "IPv4 address follows" (TBD)
> 
>      H0-DSP = "domain specific part", whose coding is determined
>               by the authority identified in the IDP.  What I show above
>               is a suggested format for current IP.
> 
>      ESI    = End System Identifier, most likely MAC address [is this
>               useful?  if not, add to "not used" bytes.].
> 
>      S = Selector bit, for use by end-systems (not used by network)
> 
> 
> 
> 
> IPv6 format, courtesy of draft-carpenter-ipv6-osi-00.txt:
> 
>      +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
>      |A | ISoc|                   IPv6 addr                   | S|
>      +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
> 
>               |------------------- H0-DSP --------------------|
> 
> Of course ISoc here is a different value from above, indicating "IPv6
> address follows".
> 
> [Note: I'm being a bit loose with terminology.  What I refer to as OSI NSAP
>        is what the ATM Forum calls "ATM End-System (AES) Address".  They're
>        structured the same, though, and I figure most people know or can 
>        look up, the structure of an NSAP.]
> 
> 

This is exactly the way IP address to ATM address mapping was proposed 
albeit for the case of the Peer model at the ATMF. The Peer model is in the
scope and charter of the MPOA group at the Forum.
--Ajay Joseph
Sr. Engineere
DeskNet Systems, Inc.
Armonk, NY 10504.
914-273-6232 x210