RE: Compatibility requirements

[Andrew Draper <ADRAPER@altera.com>]

At 16:45 PM 19/1/00 GMT, Paul Hoffman Wrote:
> At 03:51 AM 1/19/00 -0800, Andrew Draper wrote:
> >If the IDNS protocol specifies a canonicalisation algorithm then a
caching
> >server should perform correctly* regardless of how much (or how little)
of
> >that algorithm it has implemented.
> 
> I am not fond of this because it is making assumptions about where the 
> canonicalization is done. My strong preference is to do canonicalization
on 
> the client, but I don't know if we want to specify that (or not that) in 
> the requirements document. I'd be happy if we said "the protocol must 
> specify canonicalization and it must be done before the name is resolved",

> but I think that's too strong for this group.

I've been trying not to say that, but that's my conclusion as well.

Since that conclusion can be derived from the "Must not break DNS caching",
"Must not break DNSSEC" and "Must be able to upgrade canonicalisation
algorithm (if there is one)" requirements I wouldn't argue strongly that it
should be in the requirements document providing that the other three are.

    Andy