On Mon, Mar 2, 2009 at 8:57 AM, Robin Whittle <[email protected]> wrote:
> I am not sure of what the difference is between "normal" and "raw
> socket" applications.  Can you explain this and provide some examples
> of each?

Uh-oh...I now realize that I haven't spent enough time on this topic,
everything about raw sockets in the draft is just garbage...I'm
embarrassed, apologizing for the low quality.  Thanks for pointing out
this!

So I'll clean up the crap by adding a new section,  "Affected
applications and implications" covering the following topics

1. Raw socket, the most common I guess is the winpcap that is
accessing directly the network layer and bypassing the socket API so
that wireshark can capture the packets in detail

2. Mandatory changes to get the framework up and running
- DNS: RLOC extension needed for A records
- DHCP: new option needed to distribute the local RLOC to the clients

3. Applications that transport IP addresses in the payload and need to
transport RLOC identifiers to establish connections between RLOC
domains. Most of these applications have severe issues with NAT and if
we can reduce the usage of NAT the applications will be easier to
deploy
- SIP and SDP
- Mobile IP
- IPsec AH
- RSVP
- ICMP notifications
- other?

4. Applications that will function with FQDN but sometimes  IP
addresses are used instead, such as ping, traceroute, telnet, etc
We need a new CLI syntax to call the new RLOC socket API, perhaps
<EID:RLOC> is a suitable syntax, e.g, "ping 10.2.2.2:172.16.0.5" when
checking if host 10.2.2.2 is alive in RLOC domain 172.16.0.5

-- patte
_______________________________________________
rrg mailing list
[email protected]
http://www.irtf.org/mailman/listinfo/rrg

Reply via email to