Re: [Lsr] When is an IANA Registry Required

["Les Ginsberg (ginsberg)" <[e-mail]>]

Tony –

IMO, there is no need for registries for the first category. The WG has been alive for over 20 years, defined many new TLVs with flags fields, and I am not aware of any confusion – so if it ain’t broke don’t fix it.

But, if the WG consensus turns out to be to create registries in such cases, so be it – though it raises the question of what to do about existing flags fields which don’t have a registry (all of them currently). But I guess your “registry-on-write” policy could be used to address that - meaning no need to backtrack – just create registries when any new flag definitions come along.

I will leave it to the WG chairs as to how to determine WG consensus. Just hope we don’t have to write an RFC to define the WG policy on this. 😊

Thanx.

   Les


From: Tony Li <[e-mail]> On Behalf Of Tony Li
Sent: Wednesday, March 17, 2021 1:56 PM
To: Les Ginsberg (ginsberg) <[e-mail]>
Cc: Alvaro Retana <[e-mail]>; [e-mail]; [e-mail]; John Scudder <[e-mail]>; Christian Hopps <[e-mail]>; [e-mail]
Subject: Re: [Lsr] When is an IANA Registry Required


Les,


[Les:] The question here is whether there is a qualitative difference between two classes of bit fields.


That is indeed the key question. IMHO, there is not.

I don’t much care if a field is updated by a bis document or a related document. Regardless of the cause, as soon as there is more than one source of truth about the field, we are
creating ambiguity and confusion.

At the same time, I see no point in a registry with contents that never change. Thus, I will propose an alternative: by analogy to copy-on-write shared memory semantics, I propose that
we apply ‘registry-on-write’ semantics.

Specifics: When a potentially shared field is created, the defining document speciifies the name of a future registry, but does NOT request IANA create the registry at this time. When any document wishes
to update the field, it requests that IANA create it and populate it with both legacy and updated values.

Implementors that come along either document know where the source of truth is.  If the registry has not been created, then there is no ambiguity. If it has been, then there is no ambiguity.

Thoughts?

Tony