I’m a little confused by this:

> I support publication of this document, but I agree that some things
> should be improved before its ready to do so. …


Perhaps this is more of a pet peeve than a to-the-point comment.

I don’t see how one can be in support of publishing a document and then suggest 
changes.

In the case of this document, I think it needs a lot of changes (already 
suggested by many), I think publishing it as is would not be very constructive.

Looking at the comments, I keep thinking of “what is the best way to specify a 
protocol.”  This comes up here in the discussion about whether a CDS and 
CDNSKEY, if both are present, MUST or not MUST be the same.

If you think my concern is academic, I’ll refer to the flap over the 
requirement that a set of data be signed with all algorithms in the DS and 
DNSKEY set being interpreted as “the validator must see all” even if the 
validator had a way to build a chain of trust.

The basics of protocol specification include:

A statement about the set of messages that are to be exchanged.
When causes a particular statement to be sent.
The reaction to the receipt of a particular statement.

I don’t think this version of the document does a good job of that - and my 
previous comment suggested some fixes.  Whether or not the particulars of what 
I wrote are adopted, I encourage structuring the document according to the 
three lines above.

One reason software is so badly implemented is that the specifications are 
poorly written.  "One reason" - for sure.  If you ask me (and you didn’t) it is 
the responsibility of the IETF to generate quality documents and effort should 
be extended to make sure what is produced is clear and every requirement 
testable.

When it comes to saying there are statements A and B and that the two are 
supposed to be equivalent - require that.  Require (MUST) the sender to make 
the two be equivalent.  Then write rules for the receiver to deal with “what 
happens when they are not equivalent.”  The latter is done to make the protocol 
exchange more robust.

“Be conservative in what you send, be liberal in what you accept” is advice for 
implementers, not necessarily protocol designers.  The role of the latter is to 
restrict communications into a form that makes sense.


_______________________________________________
DNSOP mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dnsop

Reply via email to