Re: DNS Query Format

David MacQuigg wrote:

> > One of the 'ground rules' of SPF (current versions) is that is should not
> > require any change to the existing DNS protocols.
>
> Then SPF is already sunk.  It requires a new resource record.
>
> > Now there is nowhere within
> > the current _protocol_ in which you could carry the IP address 
> > information
> > required, so one has to resort to a 'trick'.
> I would be very surprised if the designers of DNS did not anticipate the 
> need for new and different information to be added to a query.  Looks to 
> me like they even have an "additional information" field for that 
> purpose.  Please tell me where I find any prohibition on adding 
> additional information to a query.
Oh, they did think about future record types. section 3.6 in RFC1035 
deals with the how and when it should be done.
It is not a technical reason why SPF should not require new 
infrastructure, but a business one. Time-to-market would be much slower 
if it had all types of requirements.
Also, since the effectiveness of SPF cannot be empirically demonstrated, 
it would be a hard uphill struggle to get DNS changes standardized, and 
even a tougher struggle to get them implemented.
So that's the theory, the new RRs can be added. The reality is that it's 
so close to impossible, that 'impossible' becomes a good approximation.
The last revision to DNS was in 1987, 18 years ago. So the fact that the 
DNS standard has been through the boom and bust of .com, when the 
millions of dollars were flying everywhere and there was a new 
.com-related protocol under every pile of horse-poop, is telling that 
it's not easy.
I thought this requirement of not messing with the existing 
infrastructure was one of the best decisions concerning SPF.
Radu.