Re: [dtn-interest] Late Binding examples

[Elwyn Davies <[e-mail]>]

I believe that a much earlier, IP related example, was the Nimrod 
routing and addressing scheme championed by Noël Chiappa in the early 
1990's.  This was in the 'long list' for IPng (later known as IPv6) but 
was rejected as 'still too researchy' when the short list was formed.  
This was a map based scheme whereby packets were forwarded towards their 
destination based on a 'low resolution' map that identified a gateway 
where a higher resolution map could be used to produce more closely 
focussed identification.  This process could be repeated as many times 
as required.  Considerable progress was made both theoretically and 
towards an implementation, but work was eventually abandoned before 
completion after Nimrod was rejected as an IPng candidate. I believe 
this is a form of late binding. The routing decision process was 
intended to be based on abstract maps exchanged between nodes.

The Nimrod ideas were also fed into the PNNI protocol designed for ATM 
networks.  This was implemented fully but has never been seriously 
deployed because large scale ATM never came to pass.  It used long names 
and an abstract map exchange based approach. 

There are more details and references in sections 3.3 and 3.4 of 
http://tools.ietf.org/html/draft-irtf-routing-history-10.

PNNI in particular was (IMO) flawed because it ultimately relied (as 
does the IP routing system) on a single top level authority in some form 
- there is a single tree structure at least in the 
authorization/authentication of the maps.  In a future DTN network that 
should not be the case.  I don't believe that Nimrod was intended to go 
that way either.

Joel Halpern who is still around the IETF was closely involved in the 
design and implementation of PNNI.  I have talked with him briefly about 
PNNI from an historical perspective but he may have some additional 
insights that could be instructive.

Regards,
Elwyn


Burleigh, Scott C wrote:
> ION certainly uses late binding routinely.  At every hop on the end-to-end path, the routing algorithm selects a neighboring node to transmit to and then determines what convergence-layer endpoint (e.g., UDP port number, LTP span, etc.) to transmit to in order to get the bundle to that node.  The routing is really orthogonal -- and complementary -- to late binding, which is a very late step in the forwarding procedure.
>
> Scott
>
>   
>> -----Original Message-----
>> From: [e-mail] [mailto:dtn-
>> [e-mail]] On Behalf Of Ivancic,
>> William D. (GRC-RHN0)
>> Sent: Monday, March 16, 2009 9:27 PM
>> To: dtn interest
>> Subject: [dtn-interest] Late Binding examples
>>
>>
>> One sort-of feature of DTN is late binding.  Has anyone implemented late-
>> binding (not to be confused with content-base routing/delivery)?  If so,
>> can you point me to some examples and what addressing scheme was used? I
>> ask because I am unaware of any implemented routing schemes that really do
>> late-binding.  This all is in relation to naming and addressing.  It seems
>> to me that late binding requires distributed name resolvers.
>>
>> Thanks,
>>
>>
>> /Will
>>
>>