Re: [abnf-discuss] [art] Should RFC 7405 be part of STD 68 now?

[Dave Crocker <[e-mail]>]

On 12/19/2023 8:15 AM, Carsten Bormann wrote:
> On Dec 19, 2023, at 16:45, Dave Crocker <[e-mail]> wrote:
>> The issue, then, is what it will mean to say that one uses ABNF? It will sometimes mean just the original spec and it will sometimes mean both specs.
>>
>> "It will sometimes mean" is death to interoperability.
> You are in essence arguing that a spec can never be extended.
> I don’t think this is tenable.

Your assessment of what I'm arguing is wrong.

I raised an issue.  I haven't stated my own view of what the outcome 
should be.

Changes like these never involve only the simple, focused points that 
justify advocacy.  We need to look for relevant concerns and decide 
whether they have substance, and what the tradeoffs are.


> I just checked, 17 published RFCs (beyond RFC 7405 itself) use %s”…”, in 249 lines of ABNF.
> Among these heavyweights such as RFC 9110 (HTTP semantics) and RFC 9112 (the text-based variant of HTTP).

This says that the enhancement is used.  But you did not, for example, 
compare this number against the total number of RFCs using ABNF.  Or is 
the 401 count, you listed at the end of your note, the 'total' number we 
should compare against?

And while percentage of specs is probably a useful indicator, something 
closer to percentage of actual use would be much better, of course.

Note that I did not challenge the original assertion that the 
enhancement has widespread adoption.

The question is whether it is reasonable to include it as part of the 
(mandatory) ABNF core?  This is what I believe adding it to the STD 
accomplishes, semantically.


> At this stage, I don’t think you can say you have a useful(*) ABNF implementation if you don’t implement RFC 7405.

Nicely to the point.

I think the null hypothesis form is: A modern ABNF implementation does 
not (typically) need the enhancement.

That, really, goes to the question I asked. (But I put it as a question, 
since I don't know the answer; I have nowhere near enough contact with 
the current range of ABNF implementations or usage.)



> (*)
> Yes, a total of 401 RFC reference RFC 5234 (not all of these actually contain actual ABNF lines, but this was too hard to check right now).



-- 
Dave Crocker
[e-mail]
mast:@[e-mail]
408.329.0791

Volunteer, Silicon Valley Chapter
Information & Planning Coordinator
American Red Cross
[e-mail]