All,
The requirement to have two implementations before submitting to IESG (not some "commitment to implement" before taking over a work) may be coming from two different reasons:
Reason 1 - To make sure what we are about to standardize is actually working. In many cases only during protocol implementation and testing we find issues which can be easily fixed before the document gets an RFC stamp. That is mainly the reason for IDR endorsing such a position.
Reason 2 - To make sure what we are to work on will be actually used.
It smells like the discussion at hand is about Reason 2 not Reason 1.
And let me observe that implementation does not mean that something is shipping in a product and will be accepted as a commit by the community of a given open source codebase. Many times what is implemented and tested in the interops or hackathons is a private dev which may never see a day light.
Thx,
Robert
Robert
On Mon, May 20, 2024 at 9:57 PM Joel Halpern <jmh.direct@xxxxxxxxxxxxxxx> wrote:
I was not commenting on whether the requirement was reasonable or
appropriate for email standardization. I was reacting to the assertion
that the IETF doesn't do that. The IETF does sometimes have such
requirements. Whether it should have it in the email case, and whether
it should be in the charter or somewhere else if it is appropriate, is
not something I have enough information on which to have an informed
opinion.
Yours,
Joel
On 5/20/2024 3:08 PM, Dave Crocker wrote:
> On 5/20/2024 11:58 AM, Joel Halpern wrote:
>> You do know that the IDR working group does not release any
>> specification for IETF last call without two running (and usually
>> interoperating) implementations. And everyone knows going in that
>> will be the bar to get out.
>
> Joel,
>
> 1. Perhaps you missed my acknowledgement that it is sometimes
> reasonable to impose a Proposed requirement for having interoperable
> implementations. So that's not the issue here.
>
> 2. Perhaps I didn't read carefully enough, but I do not see the
> requirement you cite stated in the IDR working group charter. So it
> might be operational practice, there, but it does not seem to have
> made it into the formal requirements, which is an issue with the
> mailmaint wg charter.
>
> 3. Given the nature of IDR technology, the importance and danger of
> IDR use, and the history of IDR problems in the operational Internet,
> the practice of requiring some interoperability testing before Last
> Call seems frankly modest to me...
>
> All that said, let me stress that unfortunately, your note seems
> wholly irrelevant to the concerns I've expressed.
>
> Both your note and Pete's suggests a really basic missing the point.
>
> d/
>
