ietf
[Top] [All Lists]

Re: An alternative to TCP (part 1)

2001-02-07 00:00:02
General Question:

- How to migrate?
 I have been using the "QWERTY" keyboard all the time.
 I am aware that it is not the best, but I am stuck.
So a set of APIs must be defined. They should be downward-compatible with basic 
socket APIs supporting TCP.

Because ATP is optimzed for IPv6, and because generally speaking a network 
application has to be recompiled if not rewritten to be migrated to IPv6 (from 
TCP/IPv4), I think it can be rationally assumed that a network application may 
be recompiled and relinked with an ATP/IPv6 library.


4. Each time the client end initiates a new connection it 
   will allocate a new IPv6 address. The allocation may 
   be done randomly, providing client anonymity.

How are you going to route those random packets?

Well, actually only the interface ID part of the IPv6 address is randomly 
chosen. The site level aggregation ID and above might be fixed or might be 
randomly chosen from a small set if the site is multi-homed. Routing of IPv6 
packet is based on hierarchical aggregation ID. The real problem is not routing 
but instead the scale problem of the network edge router: as each new 
connection consumes a new IPv6 address the router must create a new entry 
mapping the network-layer address to the link-layer address or else the ND 
protocol has to be applied again and again. A similar problem challenges the 
NAT devices as well.