Re: static versus dynamic nature of DNS: rate limiting

On Mon, Feb 23, 2004 at 04:47:15AM +0100, Patrik Fältström wrote:
> On 2004-02-22, at 22.01, Peter J. Holzer wrote:
> 
> >A new RR is a change in the protocol
> 
> No, it is not. A new RR is a change in the implementations of the 
> clients which is to be able to parse it, and the servers which are to 
> provision it. Caches does not have to be changed as it is not a change 
> in the protocol.


The DNS group told me otherwise.

In the first versions of RMX I used domain name compression to 
keep the records small and compact. Domain name compression is 
part of the DNS protocol.

Some people from the DNS group told me where I made a mistake:
If a cache does not know about these new record type and just forwards 
the reply, the reply will in most cases be broken, because the 
cache's reply will not always look exactly the same like the 
authoritative server's reply. But if not, uncompressing the reply
will not work anymore.

You can use domain name compression only if all caches can handle
the new record type. But name compression is important to keep the 
record size small. In my eyes, DNS is broken by design.

Inventing a new RR with making use of domain name compression (which
is after all part of DNS protocol) *is* a protocol change and will
break old caches.

regards
Hadmut