Re: [idn] Reality Check

[John C Klensin <klensin@research.att.com>]

--On Monday, 16 July, 2001 09:03 -0500 "Eric A. Hall"
<ehall@ehsco.com> wrote:

> 
>> It is difficult for me to imagine how you would add anything
>> out of band to finger, whois, or other "one connection, one
>> exchange" protoocols.  What did you have in mind?
> 
> Using transport-layer options to negotiate over compression,
> encryption, charset and other per-session characteristics.
>...

Ok, that is fair, although I personally believe that moving
applications options (such as most of the above) down to the
transport layer would introduce layering problems that we would
come to very seriously regret in the future.
  
> With finger, for example, the client and server could use TCP
> options to negotiate over the charset, without having to change
> the finger protocol semantics whatsoever.

Yep.  Although I think that institutionalizing that type of layer
violation would be much more likely to wreck the Internet than
anything we could do with the DNS.  Remember that the corollary
to the end to end principle that we don't change the transport
core to accomodate particular applications is considered by many
to be the key doctrine that has preserved our ability to quickly
innovate and test new ideas.

Just my impression, however.

     john