Re: [Rats] New RATS Architecture document

[Carsten Bormann <[e-mail]>]

On Oct 8, 2019, at 01:21, Henk Birkholz <[e-mail]> wrote:
> 
> Hm.. if I use requirement notation and not include the BCP14 boilerplate, that is an idnit, because you have to include the boilerplate then. And BCP14 is referring to standards track documents and not informational documents.
> 
> I've never questioned that line of thought before, tbh. There are maybe some individuals subscribed that can clarify this easily?

These individuals are called ADs :-)

I don’t think the way the IETF handles this is particularly logical, so it doesn’t quite work to make arguments entirely based on logic here.  The really important question is how is the document intended to be used.

For something like an architecture document, the document will not be used standalone, but in conjunction with a protocol specification that provides the details.  If the statements in the architecture document are necessary components to correctly interpret the protocol specification, the latter needs to make a normative reference to the architecture document.  If the architecture document is just an informational document, it needs to be blessed to be on the “downref list” before this can be done.

I don’t understand why we are creating informational documents that MUST be put on the downref list before they can be used.  But, apparently, we do, all the time.  I think it is more appropriate to go for standards track if the intention is indeed normative as a component of the protocol specification.

What is the difference?  Standards-Track documents need more IESG votes (and thus more IESG work!) than informational documents.  Putting a document that contains components of a protocol specification through the light-weight informational process and then go for downref right away sounds sneaky to this process connoisseur.  But, apparently, it can be done.

Now ask the AD :-)

Grüße, Carsten


> 
> On 08.10.19 01:10, Schönwälder, Jürgen wrote:
>> On Tue, Oct 08, 2019 at 12:41:40AM +0200, Henk Birkholz wrote:
>>> 
>>> A document that uses normative requirements terminology is standards track.
>>> 
>> I just searched backwards for the most recent informational document
>> and I found RFC (8649) and it violates this rule. RFC 7921 is
>> informational and has "Architecture" in its title and uses RFC 2119
>> language as well. So if this rule exists, the IETF is doing pretty
>> bad following it. So where is the the rule defined?
>> The question whether a document is standards-track should follow from
>> its content (is there material that is normative for creating
>> interoperable implementations?) and not from specific keywords used to
>> write this content down (or other formal aspects).
>> /js
>