Re: [ippm] Last Call: draft-ietf-ippm-twamp-session-cntrl (Individual Session Control Feature for TWAMP) to Proposed Standard

[Jerome Benoit <[e-mail]>]

Le Sun, 21 Mar 2010 10:48:36 -0700,
Murtaza Chiba <[e-mail]> a écrit :

Hello, 

> Furthermore, if the Client assigns a SID it is not guaranteed to be
> unique in the network. If the server assigns one, at the very least
> all the clients that the server accepts connections from are
> guaranteed to have unique identifiers.

I lurk this IETF WG since some time in order to enhance an old
measurement system (that need less naive measurements and hackers) used
by end-user to troubleshoot their connexion in France. Quiet honestly,
neither OWAMP or TWAMP can be implemented without a revamp to serve our
need. 

To be more explicit : 

- We can't trust the end-user clock for timestamping measurement, we
rely on time difference from the server clock source only (we plan to
calibrate timestamping inaccuracy); 
- The measurement agent run behind the end-user CPE, 1-to-n NATed for
IPv4 and firewalled, so measurement agent must start the probing
session and initiate the connection, same for probing configuration
fetching). 

The end-to-end delay and delay between landmarks and end host is a very
useful metric to do troubleshooting but we can't do it by strictly
following IETF RFC ... 

For measurement agent UUID handling, we plan to : 

- generate it at first agent start; 
- register it on the server that will handle the birthday paradox; 
- UUID has nothing to do with your probing session ID but it's the
quickest way to map a probing configuration to an measurement agent
on network topology criteria or other ... 

Maybe it will be useful to this WG. 

Cheers.

-- 
Jérôme Benoit aka fraggle
La Météo du Net - http://grenouille.com
OpenPGP Key ID : 9FE9161D
Key fingerprint : 9CA4 0249 AF57 A35B 34B3 AC15 FAA0 CB50 9FE9 161D