Re: Extensibility methods in APPCORE protocol

At 21:35 22/03/99 +0100, Jacob Palme wrote:
>At 22.55 +0100 99-03-17, Graham Klyne wrote:
>> At 13:17 17/03/99 +0100, Jacob Palme wrote:
>> >At the APPCORE meeting in Minneapolis, it was said that APPCORE should
>> >not contain any building block which is not used in two existing
>> >IETF standard.
>> I don't think extensibility is excluded by this rule -- both HTTP and SMTP
>> have extension mechanisms defined.
>Yes, but none of them has a facility for marking extensions
>as critical or non-critical, and this is something I would
>very much like to have in APPLCORE.

Well, yes, but...

In the case of HTTP I believe that there does exist such a mechanism (but I
could be wrong there).

In the case of SMTP, because the necessary information is declared up front
(by EHLO response) before a transaction is initiated, the client can base a
request upon this information so I don't see a critical/nom-critical
designation is needed.

The fact that it is not needed in all cases suggests to me that
critical/non-critical marking should not be part of a *core* protocol.


Graham Klyne

Received on Monday, 22 March 1999 16:52:38 UTC