Re: [dnsext] Clarifying the mandatory algorithm rules

[Samuel Weiler <[e-mail]>]

On Thu, 18 Nov 2010, W.C.A. Wijngaards wrote:

> Since owners MUST sign their zone correctly, I see no need to 
> specify a MUST accept missing signatures if you implement the other 
> algorithm? Why do you want to allow inproperly-signed zones?  It 
> will break resolvers that implement only a subset of algorithms?

Honestly, I haven't thought about the constraint in this light.

Some hisorical perspective may help.

Before these mandatory algorithm rules were adopted, a "properly 
signed" zone required that each RRset be signed by ANY of the DNSKEYs 
in the zone, with no regard to algorithms.  Some RRsets could be 
signed only with alg 5, some only with alg 234, and some with both.

We adopted these mandatory algorithm rules to deal with a particular 
downgrade attack (or, depending on how you look at it, a debilitating 
incentive against deploying a new algorithm).  For this purpose, it 
was (and is) sufficient to require a zone to follow the rules.  It was 
(and is) not necessary to require the resolver to check that the zone 
had followed the rules.


So we're starting from different places: I still remember the 
arguments from 2003 about why we should have ANY restrictions.  I also 
remember talking about whether we wanted to address other downgrade 
possibilities (e.g. key length) -- the WG decided not to tackle any of 
that.

-- Sam