Re: [idn] canonicalization

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

I agreed with Harald. The purpose is to decide if we wish to even do
canonicalization at all, but not how or where it is done. At least not at this
moment...

Correct me if I am wrong, I think we have consensus that *some*
canonicalization has to be done. :)

If that is the case...

Harald Tveit Alvestrand wrote:
> - The same octets in a query MUST always return the same data
>    (modulo the DNS games we all know and love :-()

I suggest not to include this as it is already described in [3].
 
> - A proposal SHOULD adhere as closely as possible to the Principle of
>    Least Astonhsiment

Cool. Do you want to define "Principle of Least Astonishment"? 

How about this.

- The protocol SHOULD have some form of canonicalization and folding rules
  such that equivalent names are folded into one. It is understood that
  these rules would not be 100% perfect or accurate.
 
> - A proposal MUST allow any client to interoperate correctly with any
>    server, as long as they implement the proposed protocol correctly

This boils down to "The protocol MUST work." :P Can we leave this out?
 
> - A proposal MUST describe how clients written to an early version of the
>    proposed protocol will interact with servers written to a later version
>    of the protocol where canonicalization is concerned; one possible means
>    is to declare any mandatory canonicalization in the client to be fixed
>    for all time at protocol release time.

Great. :-)

-James Seng