Re: [tcpm] another review of draft-ietf-tcpm-tcpsecure[-10]

[Lars Eggert <[e-mail]>]

Hi,

(individual hat on)

On 2008-9-29, at 20:31, ext Joe Touch wrote:
> Is it possible that a SHOULD isn't considered a "updates"?

RFC2119 language has nothing to do with the "updates" relationship.  
It's the applicability that matters.

"Updates 793" on document X means "if you implement 793 you also need  
to implement X". That is, because "updates" is typically used when a  
new document fixes critical bugs in an existing specification or adds  
mandatory new functionality.

For this document the applicability statement is basically "SHOULD  
implement when vulnerable, MAY otherwise". That's a conditional  
"updates". Unfortunately, the "updates" header isn't expressive enough  
to convey this.

We're left with two less-than-optimal choices: Add "updates 793",  
which doesn't capture the conditional expressed by the applicability  
statement, or omit "updates 793", which is also inaccurate.

My reason for suggesting to omit the "updates" - and this is a  
personal preference, I'll respect the WG decision and would like to  
head other opinions - is that a plain "updates 793" looks like a "you  
must implement this when implementing 793", and we had a long  
discussion around the applicability statement that seemed to indicate  
that the WG didn't want to make this blanket recommendation.

Lars

PS: As an example, look at the recent RFC5348, which is TFRC. It  
obsoletes RFC3448 (which was the old version of TFRC) but it also  
updates RFC4342 (DCCP's CCID3), which implements TFRC inside DCCP. The  
"updates" relationship here means that CCID3 implementations must now  
(also) implement RFC5348, i.e., the new version of TFRC.

_______________________________________________
tcpm mailing list
[e-mail]
https://www.ietf.org/mailman/listinfo/tcpm