[dns-operations] compressing DNS traffic data

Jim Reid jim at rfc1035.com
Thu Dec 9 10:06:09 UTC 2010

On 9 Dec 2010, at 07:48, Ray Bellis wrote:

> Strictly speaking most people have been talking about data  
> _reduction_ rather than data compression, i.e. throwing  
> "uninteresting" data away.  Hence decompression is not an issue.

Fair enough Ray. But the idea of "uninteresting" data is troubling  
too. If such data exists, it means there's excessive or redundant data  
in the DNS protocol. That seems unlikely. Even if the protocol suffers  
from excessive or redundant data, who's to say that what we consider  
"uninteresting" today becomes important tomorrow?

Some things like the source port number or query ID or QCLASS might  
appear to be attractive targets that could be discarded. I disagree.  
These things have their uses, even for traffic analysis: eg tracking  
the deployment of a resolver implementation which uses a deterministic  
set of port numbers or query IDs and the impact(s) this implementation  
has. Some might think dropping CLASS from queries and responses is a  
no-brainer. But on another thread on this list there's been a  
discussion about the problems caused by an implementation which used  
the wrong CLASS ID.

More information about the dns-operations mailing list