[Idr] Re: [Ext] Implementation requirements embedded in bgp-car [was: Re: Roman Danyliw's Discuss on draft-ietf-idr-bgp-car-14: (with DISCUSS and COMMENT)]

[Amanda Baber <[e-mail]>]

Hi John, all,

 

A note about this: when IANA is asked to use the RFC 7120 process for Specification Required registries, we ask for expert approval as well as chair and AD approval. (7120 doesn’t mention this, but 7120bis-00 does.) The expert is going to have to review the request during IETF Last Call, so we’re trying to avoid a situation where they have to say, “I wouldn’t have agreed to this, but it’s already in the registry.”

 

Would it be clear enough that the expert doesn’t have to check for two implementations if it’s an RFC 7120 early allocation request?

 

Thanks,

Amanda   

 

From: John Scudder <jgs=[e-mail]>
Date: Thursday, February 20, 2025 at 10:56 AM
To: idr chairs <[e-mail]>, "[e-mail]" <[e-mail]>, "[e-mail]" <[e-mail]>
Cc: The IESG <[e-mail]>, "Dhananjaya Rao (dhrao)" <[e-mail]>, Bruno Decraene <[e-mail]>, Clarence Filsfils <[e-mail]>, "luay.jalil" <[e-mail]>, "yitai.syc" <[e-mail]>, "[e-mail]" <[e-mail]>, James Guichard <[e-mail]>, Ketan Talaulikar <[e-mail]>, Keyur Patel <[e-mail]>, "[e-mail]" <[e-mail]>, "[e-mail]" <[e-mail]>, Hares Susan <[e-mail]>, Roman Danyliw <[e-mail]>
Subject: [Ext] Implementation requirements embedded in bgp-car [was: Re: Roman Danyliw's Discuss on draft-ietf-idr-bgp-car-14: (with DISCUSS and COMMENT)]

 

Hi All, 

 

As part of the discussion of Roman’s DISCUSS point at today’s IESG meeting, Roman raised an additional concern. Sections 10.4.1 and 10.4.2 include the requirement that the Designated Expert "Check IDR implementation report for two implementations”. Roman pointed out that the document category is Experimental, the implication being that a two-implementation requirement to obtain a code point for an experiment seems excessive.

 

Now that I come to think of it, the requirement even feels a little excessive if the document were Standards Track. It puts the applicant for the code point in a bind: it’s often hard to get implementations written if we don’t have code points. But the requirement in this document says we can’t have a code point unless there are two implementations. 

 

Maybe part of the answer to this concern is that we have RFC 7120 — it allows early allocation from Specification Required registries. Presumably, the “two implementations” rule would not be applied during the early allocation process any more than it is for other registries. (Because RFC 7120 only applies to Specification Required "where an RFC is used as the stable reference”, for any non-RFC track specification, Early Allocation wouldn’t be available, and the DE would be obliged to insist on two implementations.)

 

That leaves the residual question of whether the IDR two-implementation rule is appropriate for Experimental documents. I’m not aware of IDR having ever grappled with that question; the wiki [1] only says, "IDR generally requires at least two interoperable implementations of a draft before it is advanced to RFC”. 

 

I tend to think it is OK for IDR to require two implementations for even an Experimental RFC, but you should be aware that if you do this, you are creating a precedent, so please be sure it’s what you want. If you don’t want to create this precedent, you could remain silent (remove those two bullet points).

 

I’ll wait for a go-ahead from one of the chairs before proceeding with publication to ensure this question can be considered. (We still are waiting for Roman to clear his DISCUSS in any case.)

 

Thanks,

 

—John

 

[1] https://wiki.ietf.org/group/idr [wiki.ietf.org]