Re: Compatibility requirements

[James Seng <jseng@pobox.org.sg>]

Paul Hoffman / IMC wrote:
> >If the canonicalization is done at the client, then it will break the
> >requirement "The same name resolution request should generate the same
> >response, regardless of the location (or localisation settings) of the
> >resolver, the master server and any slave or caching servers involved."
> 
> I do not see this.
> 
> >Logic is pretty obvious.
> 
> I must be being dense. Can you give an example of what you are saying? That
> is, show an internationalized string, its canonicalized string, and how the
> request for that canonicalized string could get a different response at a
> different location of the resolver or the other servers?

Ok, lets see...

Required to have same response to same request etc. 

Now, if canonicalized is done on client before sending out, then what happened
is that different client may have different canonicalization algo and thus may
get different results.

Fallacy 1. 
Canonicalization algo need not be done on client. *doh*

Fallacy 2. 
Even if canonicalization is done on client, there is many ways an
implementation can ensure canonicalization algo is updated on the client in
some way. 

Fallacy 3.
If canonicalization is done on the client before request is send out, and if
two client has different canonicalization algo, this means the request which
is been send out are *different* and thus it never breaks the first
requirement. 

The interesting part is Fallacy 3. It occurs to me that canonicalization on
the client may have an effect to the overall operations of IDN. So afterall,
we may want to define how the client behave afterall. 

-James Seng