Re: empty lists, was: Benjamin Kaduk's Discuss on draft-ietf-httpbis-header-structure-18: (with DISCUSS and COMMENT)

[Julian Reschke <[e-mail]>]

On 19.05.2020 14:20, Mark Nottingham wrote:
> On 19 May 2020, at 8:04 pm, Julian Reschke <[e-mail]> wrote:
>>
>> It really doesn't help a lot with the case of:
>>
>>   Foo:
>>
>> What's the point in disallowing this in the ABNF (and on the wire) when
>> the normative parsing algorithm will accept it?
>
> As has been said many times, the ABNF in the specification is not normative; it's illustrative. The proposed edit clarifies that. If that's still felt to be confusing, the right thing to do would be to remove ABNF from the spec completely, to avoid the confusion. I'm happy to do that if the IESG wishes so.

I get that. What I'm saying is that this is confusing, and that there
would be less confusion if the ABNF simply allowed empty lists.

WRT removing the ABNF: please, no. Absent the ABNF, the algorithms are
the only thing to look at, and they really only help if you implement them.

> Julian, you've had many opportunities to bring this up before (having participated in the original issue about empty lists). It would have been much more helpful if you'd expressed your concerns earlier -- even during WGLC -- instead of after IETF LC.

You might recall that during WGLC I pointed out that it's extremely hard
to review the algorithms without actually implementing them, that's why
I wasn't aware of the issue
(<https://lists.w3.org/Archives/Public/ietf-http-wg/2020JanMar/0138.html>).

I now actually *did* implement the spec so to be able to properly review
it. So please don't shoot the messenger, in particular if the feedback
comes just a few days after end of LC (which, FWIW and as a reminder,
was not announced on the WG mailing list).

Best regards, Julian