[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Architectural approaches to multi6

On dinsdag, maa 25, 2003, at 17:14 Europe/Amsterdam, Brian E Carpenter wrote:

3: remove addresses from all places in the protocol stack above the IP
layer. If higher layers are unaware of addresses and the addresses
follow the topology, IP can route around failures by changing
addresses.


We've been preaching against exposing addresses above the transport
layer since RFC 1900 (dated 1996) at least. Little has changed in the
real world.




As far as I can tell, most people/applications use names rather than 
addresses most of the time. The trouble is that protocols such as TCP 
use the addresses in the IP header in their processing. If we make 
these protocols look at something else (the FQDN, an address that's not 
in the IP header, ...) we are free to change addresses at any time 
without breaking sessions.




And I have reason to believe transition won't be as hard as it may 
seem

at first glance.




Not sure I understand that statement.




Changing all references to addresses everywhere except in the IP layer 
is a huge undertaking. However, I think multihoming benefits can be 
achieved before this process has been completed.