We're not exactly sure what "we cannot have a normative reference from the
1.1 specification to the 1.0 specification" means.
Our current IPP/1.1 Encoding and Transport I-D contains the following
statements which refer to the experimental RFCs: 2565 and 2566 using a
SHOULD, not a MUST:
IPP/1.1 server implementations SHOULD interoperate with IPP/1.0 client
implementations, as defined in [rfc 2565] and [rfc 2566] documents. If an
IPP/1.1 server implementation does not support an IPP/1.0 client, it MUST
return the error 'server-error-version-not-supported' and the version in the
response MUST be a version that the server supports and SHOULD be a version
that is closest to the clients version in the request.
Is this a "normative" reference or not?
In other words, is this ok for our IPP/1.1 document or will "we need to
request a variance from the normal rule from the IESG" as you mention below.
From: Manros, Carl-Uno B [mailto:firstname.lastname@example.org]
Sent: Friday, May 14, 1999 17:03
Subject: IPP> FW: FW: ADM - Questions on Backwards Compatibility between
IPP/1. 0 an d IPP/1.1
I have now got a response from Keith Moore on our earlier question about
referencing IPP1.0 in the IPP/1.1 document. See his thoughts below.
When he mentions "Last Call" I believe he means the IETF wide Last Call
which is issued by the IESG after we have given them the documents.
Also, it now looks unclear whether Keith will be able to dial in to
our meeting in Philadelphia - please stay tuned.
From: Keith Moore [mailto:email@example.com]
Sent: Friday, May 14, 1999 3:17 PM
To: Manros, Carl-Uno B
Cc: Moore, Keith; Manros, Carl-Uno
Subject: Re: FW: ADM - Questions on Backwards Compatibility between
IPP/1.0 an d IPP/1.1
sorry, I thought I answered this.
in general, we cannot have a normative reference from the 1.1 specification
to the 1.0 specification.
an alternative would be to incorporate the relevant parts of the 1.0
specification into the 1.1 specification, say as an appendix.
another alternative would be to request a variance from the normal rule
forbidding normative reference to informational/experimental specifications.
we might be able to do this on the same Last Call as for IPP 1.1.
either way, the biggest trick is probably authentication - if you require
implementations to support 1.0 then they will have to be able to either
disable 1.0 support, or support the same authentication methods for 1.0
that they do for 1.1. (in other words, we can't open up a security
hole by requiring support for 1.0)
p.s. I have just found out that I need to be in DC on May 27, so I don't
know whether I'll be able to join the Philadelphia meeting by telephone.
Will let you know what works out.