RE: [idn] Requirements: caching servers

[Andrew Draper <ADRAPER@altera.com>]

On 02 Feb 00 Harald Tveit Alvestrand wrote:
> Agree to what we want to achieve.
> I've got trouble with the word "understand" here. Possibly the requirement

> we want to capture on caching servers is:
> 
> <start propsed text>
> A caching server MUST NOT return data in response to a query that would
> not have been returned if the same query had been presented to the origin
> server. This applies fully for the cases when:
> 
> - The caching server does not know about iDNS (that is, old servers)
> - The caching server implements the whole specification
> - The caching server implements a legal subset of the specification
> <end proposed text>

Agree.  This is pretty much what my original confused text was trying to
say, although I was also thinking of the effect that upgrades to ISO10646 or
the canonicalisation algorithm would have.

I would change the third case to:

| - The caching servers implements a legal subset of the specification
|   (for example an older IDN version than the authoritative server)

> It's not totally unthinkable that there could be a time when:
> 
> - An origin server implementing an iDNS would return the same answer for
>    queries on "A" and "a" (A-ring and a-ring)
> - A cache server would not know to fold case on these, so would forward
>    a query for "A" even if it had a cached entry for "a"
> - The overall system would still work, seen from the user.

I can see things working that way during the transition.  It would be nice
if the fully transitioned system did not generate the extra traffic which
will happen during the transition, but I think there is already a
requirement for that.

Regards,

    Andy