Dave,

Thanks. Pls see inline,

For what it’s worth, my Internet draft also discourages connection/destination 
logging - draft-daveor-cgn-logging-04 (see section 3).

Besides the size of the log data, the CGN implementations may take a 
performance hit if destination A+P also needs to be logged (e.g. connection 
log), resulting in increased CGN investment.


outlined the regulatory alternatives that are the only options left for dealing 
with CGN crime attribution (if source port logging at internet facing servers 
does not become routine) - one of which was this form of connection logging.

The need for connection logging may go beyond the concern of size of logging 
data - user privacy.  And this carries over to not only A+P techniques, but 
also IPv6. IOW, this concern may not be limited to address sharing techniques.

Cheers,
Rajiv Asati
Distinguished Engineer, Cisco Services


On May 3, 2018, at 8:58 PM, Dave O'Reilly 
<[email protected]<mailto:[email protected]>> wrote:

Hi Rajiv,

For what it’s worth, my Internet draft also discourages connection/destination 
logging - draft-daveor-cgn-logging-04 (see section 3).

Re the Indian government mandated connection logging that you mentioned - I was 
not aware of this but it is a piece of strong supporting evidence for exactly 
the point I was making in an email earlier this week 
(https://www.ietf.org/mail-archive/web/int-area/current/msg06442.html) where I 
outlined the regulatory alternatives that are the only options left for dealing 
with CGN crime attribution (if source port logging at internet facing servers 
does not become routine) - one of which was this form of connection logging.

As I said at the time, the crime attribution information gap introduced by CGN 
is a problem right now, and something is going to have be done about it, either 
by the “internet” (as I’m trying to advocate for), or if not, then by 
regulatory action that will be introduced in individual jurisdictions in due 
course. I reiterate the point I made at the time: when ISP regulators get their 
hands on a problem, they come up with ISP-centric solutions, all of which are 
far worse from a privacy point of view than source port logging.

I predict that many other national ISP regulators will be forced to act on this 
problem in the coming years, and in the absence of meaningful alternatives will 
mandate similar logging.

daveor


On 3 May 2018, at 22:50, Rajiv Asati (rajiva) 
<[email protected]<mailto:[email protected]>> wrote:

Is there an RFC (besides 6269) that encourages / discourages CGN logging of 
destination IP+Port if source IP+port is already logged?

RFC6269 does mention the below, as compared to the server side logging of 
source IP+port -
// logging the destination address on the NAT is inferior
  to logging the source port at the server.
https://tools.ietf.org/html/rfc6269
//

(BTW, having both source+destination in the NAT log implicitly means no bulk 
allocation of source ports possible)

Separately, this prohibits using stateless NAT based solutions such as MAP or 
using deterministic NAT, since there is no logging in such solutions. If such a 
guideline was also mandated for native IPv6, then it would pose an interesting 
deployment issue.

--
Cheers,
Rajiv Asati
Distinguished Engineer, Cisco

PS: Few may be aware of Govt. of India’s mandate* to log both source and 
destination IP+port pair.
Click on “Parameter to be stored in SYS Log of Network Address Translation 
(NAT) for Internet Access” on this page - 
https://www.corestack.io/blog/the-log-mandate-enabling-indian-isps-to-adhere-to-dot-compliance-rules/

PS:
https://tools.ietf.org/html/rfc6302
https://tools.ietf.org/html/rfc7422


Session and service continuity
_______________________________________________
Int-area mailing list
[email protected]<mailto:[email protected]>
https://www.ietf.org/mailman/listinfo/int-area

_______________________________________________
Softwires mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/softwires

Reply via email to