Re: Murray Kucherawy's Discuss on draft-ietf-httpbis-client-hints-13: (with DISCUSS and COMMENT)

On Mon, May 11, 2020 at 7:24 PM Murray S. Kucherawy <superuser@gmail.com>
wrote:

> Thanks for replying.  As a reminder, once you've dealt with the IANA
> concern, the rest of these comments are non-blocking.  However, I'm happy
> to discuss them too.
>
> On Mon, May 11, 2020 at 12:58 AM Yoav Weiss <yoav@yoav.ws> wrote:
>
>>
>> * "Implementers SHOULD be aware ..." -- this feels like an
>>> awkward construction; might I choose not to be aware?
>>>
>>
>> MUST be aware?
>>
>
> Just "need to be aware".  I find it awkward to apply
> compliance/requirements language to people.
>
> * "Such features SHOULD take
>>> into account ..." -- same issue as before, this seems an odd use of BCP
>>> 14
>>> language
>>
>> * "User agents SHOULD consider ..." -- same
>>
>> * "Implementers ought to
>>> consider ..." -- why is this only "ought to" given the prior SHOULDs?
>>>
>>
>> Would turning all those to a MUST work?
>>
>
> In the third bullet I was trying to illustrate my point: I think it makes
> more sense not to use requirements language when talking about people, so
> that one seems right to me.
>
> RFC2119 Section 6 gives guidance that's relevant here.
>
> In the first two bullets above, I don't know how to measure compliance
> with the requirement that a feature SHOULD take something into account, or
> SHOULD consider something.  I think the guidance you're providing needs to
> either be more direct and explain what compliance looks like, or not use
> these key words at all (at least not in their all-caps forms) if all you
> want to do is bring a particular topic to the attention of an implementer.
>
> You also have a "SHOULD take into account" wrapping a list that has two
> SHOULD NOTs in it.  I don't know how to interpret that.
>

Thanks for clarifying. I'll change those to "need to", as you suggested.


>
> Section 6.1:
>>> * Why does "Specification document(s)" refer to only a specific section
>>> of this
>>> document?  Isn't the whole document applicable?
>>>
>>
>> Sure. It's currently pointing at the specific section that defines the
>> header, but I can change it to refer to the whole document if that's
>> preferred.
>>
>
> The registration document says you can do it either way.  I just find
> identifying a specific section to be unusual and I was curious.
>

I don't think there was a particular reason for that, other than trying to
point folks at the relevant section.


>
> -MSK
>