On 4/1/25 12:10, Pawel Kowalik wrote:
Hi Andy,
On 31.03.25 16:50, Andrew Newton (andy) wrote:
Hi all,
At IETF 122, Pawel brought up the lack of time to discuss the simplification of the
extension rules as outlined in the email below. From what I can tell, the working group
agrees with the simplification of rules on writing RDAP extensions, with the exception of
Pawel. In fairness to him, this warrants a bit more discussion as his position, as I
understand it, is not a simple "I disagree."
As I understand it (and Pawel please correct me), his position is that
violation of the rules should be NOT RECOMMENDED whereas our statement below
implies MUST NOT.
[PK] Well, I don't like framing of my position with a lot of rhetoric. "violation of
the rules" sounds so obvious that it should be forbidden, that it frames directly
any disagreement into a difficult position to argument for it.
No offense intended, but they are rules even if there are exceptions.
In fact "the rules" set in 2.1 of RFC9083 are no rules, but a recommendation
(SHOULD) itself. So I argument actually to keep the status quo of RFC9083 as opposed to
defining new rules as now the change of -05 proposes.
I agree that a SHOULD without a qualification as to the consequences of not
following that SHOULD are not really rules. And that is the fundamental issue.
We have to qualify that SHOULD otherwise it because a MUST.
Also in the original E-mail you mentioned "complex set of rules" that hinder
interoperability without actually any evidence which these are or would be. I went though
all changes in -05 and I didn't find any point where the rules got simplified in any way.
The complex set of rules are upon the extension authors and the designated
experts approving extension registrations. By simplifying the rules for writing
extensions, we decrease the chance that an extension has interoperability
issues.
-05 does not contain the simplifications. This was a proposal I suggested after
addressing the 22 issues that got us to -05. We can produce an -06 with the
simplifications for comparison.
Finally I argument that the provisions of STD 95 are absolutely sufficient to
maintain interoperability. By including the changes of -05, even though the
document ought to guide extension authors not the implementations, it might
either misguide the implementations which would implement stricter rules and
not be able to handle extension from before extension draft publication as RFC
- so the interoperability would suffer in the end.
IMO, things like NOT RECOMMENDED and SHOULD/SHOULD NOT are nearly worthless
unless they can be qualified. That is, unless we can describe the conditions
for going against a recommendation then there is no clear need to allow doing
so. And that isn't just my opinion: the IESG routinely puts DISCUSSes on docs
for this.
[PK] That is correct and likely right to do so. Worth mentioning that the -05 document uses
"RECOMMENDED" in 9 places and "SHOULD" in 39 so I really don't take it as a
valid criteria to decide whether to change RFC9083 / STD 95 or not.
This is a miscommunication on my part, I guess. -05 does not contain the simplifications.
However, without the suggested simplification we will need to qualify most of those 9
RECOMMENDS and 39 SHOULDs making the "rules" even more difficult to comprehend.
-andy, no hat
_______________________________________________
regext mailing list -- regext@ietf.org
To unsubscribe send an email to regext-le...@ietf.org
